parisa cheraghi - epubs.surrey.ac.ukepubs.surrey.ac.uk/855173/1/27558473.pdf · parisa cheraghi...
TRANSCRIPT
Fast and Accurate Spectrum Sensing in Low Signal to Noise Ratio Environment
Parisa Cheraghi
Submitted for the Degree of Doctor of Philosophy
from the University of Surrey
4 UNIVERSITY OFm SURREY
Centre for Communication Systems Research Faculty of Engineering and Physical Sciences
University of Surrey Guildford, Surrey GU2 7XH, U.K.
September 2012
@ Parisa Cheraghi 2012
ProQuest Number: 27558473
All rights reserved
INFORMATION TO ALL USERS The qua lity of this reproduction is d e p e n d e n t upon the qua lity of the copy subm itted.
In the unlikely e ve n t that the au tho r did not send a co m p le te m anuscrip t and there are missing pages, these will be no ted . Also, if m ateria l had to be rem oved,
a no te will ind ica te the de le tion .
uestProQuest 27558473
Published by ProQuest LLO (2019). C opyrigh t of the Dissertation is held by the Author.
All rights reserved.This work is protected aga inst unauthorized copying under Title 17, United States C o de
M icroform Edition © ProQuest LLO.
ProQuest LLO.789 East Eisenhower Parkway
P.Q. Box 1346 Ann Arbor, Ml 4 81 06 - 1346
AbstractOpportunistic Spectrum Access (OSA) [1] promises tremendous gain in improving spectral efficiency. The main objective of OSA is to offer the ability of identifying and exploiting the under-utilised spectrum in an instantaneous manner in a wireless device, without any user intrusion. Hence, the initial requirement of any OSA device is the ability to perform spectrum sensing. Local narrow-band spectrum sensing has been quite well investigated in the literature. However, it is realised that existing schemes can hardly meet the requirements of a fast and accurate spectrum sensing particularly in very low signal-to-noise-ratio (SNR) range without introducing high complexity to the system. Furthermore, increase in the spectrum utilisation calls for spectrum sensing techniques that adopt an architecture to simultaneously search over multiple frequency sub-bands at a time. However, the literature of sub-band spectrum sensing is rather limited at this time. The main contributions of this thesis is two-fold:
• First a clusterd-based differential energy detection for local sensing of multicarrier based system is proposed. The proposed approach can form fast and reliable decision of spectrum availability even in very low SNR environment. The underlying initiative of the proposed scheme is applying order statistics on the clustered differential Energy Spectral Density (BSD) in order to exploit the channel frequency diversity inherent in high data-rate communications.
• Second contribution is three-fold : 1) re-defining the objective of the subband level spectrum sensing device to a model estimator, 2) deriving the optimal model selection estimator for sub-band level spectrum sensing for fixed and variable number of users along with a sub-optimal solution based on Bayesian statistical modelling and 3) proposing a practical model selection estimator with relaxed sample size constraint and limited system knowledge for sub-band spectrum sensing applications in Orthogonal Frequency-Division Multiple Access (OFDMA) systems.
The result obtained showed that through exploitation of the channel frequency selectivity the performance of the stat-of-the-art spectrum sensing techniques can be significantly improved. Furthermore, by modelling the sub-band level spectrum sensing through model estimation allows for new spectrum sensing approach. It was proved both analytically and through simulations that the proposed approach have significantly extended to state-of-the-art spectrum sensing.
K ey words: Differential, energy detection, low signal-to- noise ratio (SNR), multicarrier, opportunistic spectrum access, spectrum sensing.
Acknowledgements
Any attempt to list the people and opportunities with which my life has been richly blessed would be like trying to count the stars in the sky. Yet among these stands individuals whose profound impact deserves special acknowledgment and to whom I would like to dedicate this thesis.
First and foremost, I would like to express my gratitude to my principle supervisor, Dr. Yi Ma for constant and generous support and guidance, whom his mind provoking discussions, careful comments and criticism have greatly influenced my research. The care and time he put into all his students set an example I hope to follow.
Secondly, I would like to express my most sincere appreciations to my co-supervisor. Professor Rahim Tafazolli for his endless support throughout my Ph.D. His patient, thought-provoking guidance and instruction provided a foundation that will continue to influence my research.
I wish to express my deepest regards to my parents for their endless love and encouragement and without the support of whom, I could not have been able to reach this stage.
I would like to thank all my friends and colleagues in CCSR for their support during my Ph.D.
Last but no means least, I would like to thank all my teachers and lecturers from the first day of school up until now for their undeniable contributions towards my academic achievements.
Contents
IX
1 Introduction 1
1.1 Background................................................................................................ 1
1.2 Motivation and O b jec tiv e ....................................................................... 3
1.3 Major Contribution ................................................................................ 3
1.4 Thesis O rgan isa tion ................................................................................ 5
1.5 Publication L is t .......................................................................................... 5
1.5.1 Journal Publications..................................................................... 5
1.5.2 Conference P roceed ings............................................................... 5
1.5.3 P a t e n t s ........................................................................................... 6
2 State-of-the-art Spectrum Sensing Techniques 7
2.1 Problem Formulation for Narrow-Band Spectrum S en sin g ................. 7
2.2 Exploiting Energy D ifference................................................................. 9
2.2.1 Energy Detection........................................................................... 9
2.2.2 Energy Detection Under Noise U n ce rta in ty ............................. 10
2.3 Exploiting Stationarity Difference.......................................................... 12
2.3.1 Second Order Cyclostationarity D etection ................................ 12
2.3.2 Covariance Based D etection......................................................... 14
2.3.2.1 Eigenvalue Based D etection........................................ 16
2.3.3 Matched F i l te r in g ........................................................................ 17
2.3.3.1 Pilot Based Match Filtering D etection....................... 17
2.4 Exploiting The Distribution Difference................................................. 18
2.4.1 Entropy Based D e te c tio n ........................................................... 19
2.4.2 Kurtosis Based D e tec tio n ........................................................... 20
2.5 Summary of The Narrow-Band Spectrum Sensing Approaches . . . 21
2.6 Sub-Band Level Spectrum Spectrum S e n s in g .................................... 23
2.6.1 Sub-Band Level Spectrum Sensing System M o d e l .................. 23
ii
Contents iii
2.6.2 Filter-Bank Based Spectrum Sensing....................................... 23
2.6.3 Joint Multi-Band D etection ....................................................... 24
2.6.4 Sequential Multi-Band D etection .............................................. 25
2.6.5 Wavelet Based Detection .......................................................... 26
2.6.6 Wigner-Ville Based D etection.................................................... 27
2.7 Summary of The Sub-Band Spectrum Sensing Approaches................ 28
2.8 S u m m a ry ................................................................................................... 29
3 Cluster-Based Differential Energy D etection for Spectrum Sensingin Multi-Carrier System s 31
3.1 Introduction................................................................................................ 31
3.2 System Model and Problem Formulation.............................................. 32
3.2.1 Multi-Carrier S ystem s................................................................. 32
3.2.2 Effect of Second Order M om ent................................................. 34
3.2.3 Statement of The Spectrum Sensing P ro b le m ......................... 34
3.3 Cluster-Based Differential Energy D etection ........................................ 35
3.3.1 Sensing of CP-Based Multi-Carrier Signals ............................ 35
3.3.2 Overcoming Timing Offset........................................................... 41
3.3.3 Extension to the ZP-Based S ystem ........................................... 42
3.3.4 Knowledge of Key Parameters ................................................. 42
3.4 Performance Analysis ............................................................................. 43
3.4.1 Probability of False A larm .......................................................... 43
3.4.2 Probability of D etection ............................................................. 46
3.4.3 Numerical Results and Discussions........................................... 49
3.4.4 Computational C om plexity ....................................................... 52
3.5 Simulation Results and Discussions....................................................... 52
3.6 S u m m a ry .................................................................................................. 57
4 A Bayesian M odel Based Approach for Joint Sub-Band Level Spectrum Sensing 62
4.1 Introduction................................................................................................ 62
4.2 Problem Formulation................................................................................ 63
4.3 Optimal and Sub-Optimal S o lu tio n s .................................................... 64
4.3.1 Optimal Decision R u le ................................................................ 64
4.3.2 Optimal Solution For Fixed â .................................................... 65
4.3.3 Sub-optimal Solution................................................................... 67
4.4 Link to OFDMA based Systems............................................................. 69
Contents iv
4.4.1 Transmitted S ignal....................................................................... 70
4.4.2 Received Signal............................................................................. 70
4.4.3 Join Sub-Band Level Spectrum Sensing in OFDMA Systems 70
4.5 Performance Analysis for The Practical S o lu tio n .............................. 73
4.5.1 Asymptotic Probability of False Alarm and Threshold . . . . 73
4.5.2 Asymptotic Probability of “knee-point” D e te c tio n .................. 74
4.5.3 Optimal Asymptotic Threshold S e t t i n g .................................... 76
4.5.4 Computational complexity........................................................... 77
4.6 Further Discussions.................................................................................. 78
4.6.1 Overcoming Timing Offset........................................................... 78
4.6.2 Overcoming the Energy L ea k ag e ............................................... 78
4.6.3 Extension to the ZP-based s y s te m ........................................... 79
4.7 Simulation R e s u lts .................................................................................. 80
4.8 S u m m a ry ..................................................... ............................................ 84
5 Conclusion and Future Work 86
5.1 Conclusion .............................................................................................. 86
5.2 Future W o r k ................ 90
Appendix A 91
Appendix B 92
Appendix C 94
Appendix D 95
Appendix E 96
Bibliography 98
List of Figures
2.1 Flow chart of conventional energy detection........................................................... 9
2.2 Flow chart of frequency-domain energy detection................................................. 9
2.3 The SNR wall phenomenon.......................................................................................... 11
2.4 Flow chart of second order cyclostaionary based detection technique. . . . 12
2.5 Flow chart of covariance based detection................................................................ 14
2.6 Block diagram of Eigenvalue based detection technique..................................... 16
2.7 Block diagram of the pilot based detection technique......................................... 18
2.8 Flow chart of entropy based detection...................................................................... 19
2.9 Flow chart of kurtosis based detection..................................................................... 20
2.10 Block diagram of the wavelet based detection technique.................................... 26
3.1 Block diagram of the cluster-based differential energy detection algorithm .................................................................................................................................. 36
3.2 Effect of the sort function on the output, for iV = 50 and r = 7 on various distributions. This figure illustrates how the sort function focuses on a particular point of a distribution based on the value of r. Furthermore the shape difference for various distribution all having a mean value of 0.42 isalso shown in this figure............................................................................................... 39
3.3 The relationship between the PD and the observation length for M = 64and £ = 6 .......................................................................................................................... 49
3.4 The relationship between the PD and the coherence bandwidth, £ , andthe observation length A = 1 0 ................................................................................... 50
3.5 Complementary ROC curves of the Test I and it’s comparison with energy detection for various uncertainty factors (U), the optimal detector based on Neyman-Pearson criteria. 7 = —IQdB, £ = 8 and M = 64 based onthe analytical results in Section 3.4.......................................................................... 50
3.6 Complementary ROC curves of the Test II and it’s comparison with energy detection for various uncertainty factors (U). 7 = —IQdB, £ = 5 andM = 64 based on the analytical results in Section 3.4........................................ 51
3.7 Comparison of the simulation results and its equivalent analytical results (in Section 3.4) for Test I. Furthermore the effect of the sorting operationis shown............................................................................................................................. 53
3.8 Comparison of the simulation results of the proposed technique with andwithout the sorting operation and its equivalent analytical results (in Section 3.4) for Test II........................................................................................................ 54
V
List of Figures vi
3.9 The relationship between the PD and the observation length for M = 64and £ = 6 .......................................................................................................................... 59
3.10 The relationship between the PD and the coherence bandwidth, £ , andthe observation length K = 1 0 ................................................................................... 59
3.11 Complementary ROC curves of the Test I and it’s comparison with energy detection for various uncertainty factors (U), the optimal detector based on Neyman-Pearson criteria. 7 = —IQdB, £ = 8 and M = 64 based onthe analytical results in Section 3.4.......................................................................... 60
3.12 Complementary ROC curves of the Test II and it’s comparison with energydetection for various uncertainty factors (U). 7 = —lOdB, £ = 5 andM = 64 based on the analytical results in Section 3.4........................................ 60
3.13 The performance comparison of the proposed technique, frequency-domain energy detection, second order cyclostationarity, pilot based detection and differential energy detection for K = 7.................................................................... 61
3.14 The effect of the differential and clustering stages on the performance ofthe proposed spectrum sensing technique............................................................... 61
4.1 The performance evaluation of the optimal solution introduced in Section4.3 for fixed â and K = 100..................................................................................... 67
4.2 The performance comparison of the proposed optimal and sub-optimalsolutions derived in Section 4.3......................................... 69
4.3 Flow chart of the proposed OFDMA sub-band level spectrum sensing. . . 71
4.4 ROC Curve of the proposed algorithm in Section 4.4.3 for p = 2 dB. . . . 76
4.5 Step by step process of the proposed algorithm in Section 4.4.3 in order todetermine the vacancy of unused sub-bands.......................................................... 81
4.6 ROC curve comparison of the proposed algorithm and energy detectionbased filter-bank approach........................................................................................... 82
4.7 Performance comparison of the the proposed algorithm and energy detection based filter bank in terms of probability of detection for PFA = 0.01. 83
List of Tables
2.1 Summary of the state-of-the-art local narrow-band spectrum sensingapproaches................................................................................................... 22
2.2 Summary of the state-of-the-art local sub-band spectrum sensing approaches........................................................................................................ 29
3.1 Comparison of the state-of-the-art local narrow-band spectrum sensing approaches and the cluster-based energy detection......................... 56
4.1 Extended Pedestrian A m odel................................................................. 80
4.2 Comparison of the state-of-the-art local sub-band spectrum sensingapproaches and proposed Bayesian approach......................................... 85
VII
List of Abbreviation
OSA Opportunistic Spectrum AccessFCC Federal Communication CommitteeOfcom Office of CommunicationCSMA Carrier Sense Multiple AccessPFA Probability of False AlarmPD Probability of DetectionSNR Signal-to-Noise RatioBSD Energy Spectral DensityOFDMA Orthogonal Frequency-Division Multiple AccessOFDM Orthogonal Frequency Division MultiplexingFFT Fast Fourier TransformIFFT Inverse Fast Fourier TransformIDFT Inverse Discrete Fourier TransformAWGN Additive White Gaussian NoiseDFT Discrete Fourier TransformCP Cyclic PrefixIBI Inter-Block InterferenceCFO Carrier Frequency Offseti.i.d. Independent and Identically DistributedMGF Moment Generating Functionp.d.f. Probability Distribution FunctionCDF Cumulative Distribution FunctionMC-CDMA Multi-Carrier Code Division Multiple AccessSC-FDMA Single-Carrier Frequency Division Multiple AccessMIMO Multiple-input multiple-outputROC Receiver Operating Characteristic3GPP 3rd Generation Partnership Project (Telecommunication)LTE Long Term EvaluationPDP Power Delay ProfileZP Zero PaddingPSD Power Spectral Density
Vlll
List of Symbols
s{t) Transmitted Signalv{t) Additive White Gaussian Noisey{t) Received signal at the spectrum sensing device'Ho Hypothesis condition that the spectrum band of interest is vacant7^1 Hypothesis condition that the spectrum band of interest is occupiedA Threshold value^ Test statistic used for spectrum sensing^ s c Test statists for second order cyclostationarity based detection^CD Test statistics for covariance based detection^EDt Threshold value to time domain energy detectionN Observation lengthn Discrete time sample indext Continuous time sample index^ N x N N X N Discrete Fourier Transform Matrix<7 Additive white Gaussian noise varianceCTg Transmitted signal varianceXe d Threshold value used for energy detectionU Noise uncertainty factorSNRwaii Signal to noise ratio walli F i ( . ; .) Hypergeometric functionF(.) Gamma Function'P Signal periodRy Autocorrelation of variable yX Cyclic frequencyL n ( - ) Laguerre polynomial functionIf, L x L Identity matrixCy Covariance matrix of the received signalCs Covariance matrix of the transmitted signalRy Frequency representation of cyclic autocorrelation function of variable yQmax Maximum eigenvalue of the covariance matrixQmin Minimum eigenvalue of the covariance matrix^EV Test statistic for eigenvalue based detectionXe v Threshold value for eigenvalue based detectionâ Average eigenvalueSp{ t ) Pilot signal^PM Test statistics used for pilot based match filtering detectionApM Threshold value used for pilot based match filtering detection^ e b Test statistics used for pilot based entropy based detectionXe b Threshold value used for pilot based entropy based detectionlog Logarithm operationy,y Average value of the received signal
ix
^ k b Test statistics used for Kurtosis based detectionXk b Threshold value used for Kurtosis based detectionK Total number of sub bands^WB Test statistics used for Wavelet based detectionXwB threshold value used for Wavelet based detectionI m M X M identity matrix$ Pre-coding MatrixJ Total number of sub-carriersM Number of data Sub-carriersh Frequency selective channele Frequency offset normalised by the sub carrier spacinge Timing offsetTb Block durationTs Sampling periodAfo Noise PowerX OFDM modulated transmitted signalB Estimate of the channel coherence bandwidthC Upper bound of channel bandwidthL Cluster size[.J Floor functionIf Estimated power spectral densityà Estimated spectrum availability indicator functionQ Normalised covariance matrixdet(.) Determinant functionerfc(.) Complementary error functionA(.) Upper triangle channel matrixV(.) Lower triangle channel matrix
Central Chi squared distribution with N degrees of freedom X^(A) Chi squared distribution with N degrees of freedom with non-centrality parameter At q Values of q sorted in increasing order0(a, b) Guassian distribution with mean and variance a and b respectivelyI j J X J identity matrixC m M X M circulant channel matrixp Signal-to-noise ratio^ Matrix transpose operation* Complex conjugate operation
Hermitian function E(.) Expectation operationmax Maximum function, return the maximum value in a setmin Minimum function, returns the minimum value in a set|.| Absolute Value functionI.II Frobenius norm
Chapter 1Introduction
1.1 Background
Opportunistic spectrum access (OSA) [1], first coined by Mitola et al. [2] under
the term “spectrum pooling” in cognitive radio terminology, promises tremendous
gain in improving spectral efficiency. The main objective of OSA is to offer the
ability of identifying and exploiting the under-utilized spectrum in an instantaneous
manner in a wireless device, without any user intrusion. This allows the wireless
devices to rapidly change their modulation scheme and communication protocol
so as to better and more efficient communication. The initial requirement of any
OSA device is to determine the spectrum availability. There are three possible
solutions for monitoring the spectrum availability proposed in the literature: 1)
through an ubiquitous connection to the database, 2) a dedicated standardised
channel to broadcast a beacon signal, 3) spectrum sensing [3]. Recently, Federal
Communication Committee (FOG) [4] considered database connection for inclusion
in the IEEE 802.22 standard [5]. However, it has been shown in [6] that the geo
location database solution might incur additional costs, e.g., signalling overhead,
scheduling complexity and database maintenance costs. Nevertheless, spectrum
sensing, thanks to its relatively low infrastructure cost, still receives more and more
attention. Therefore, proposed by the Office of Communications (Ofcom) [7] a
complementary application of both spectrum sensing and database connection can
provide a practical solution for enabling spectrum availability monitoring.
Spectrum sensing is a traditional topic in the scope of signal processing for mobile
communications. It is quite mature for carrier sense multiple access (CSMA) [8]
1.1. Background
based random access networks, i.e., “listen before talk” communication protocols.
In such communication protocols the transmitter uses feedback from the receiver
in order to evaluate the availability of a channel. Hence, the transmitter has to
detect the presence of an encoded signal from another station before attempting to
transmit. If a carrier is sensed, the station waits for the transmission in progress
to finish before initiating its own transmission. New challenges and problems arise
for spectrum sensing in flexible networks, e.g., cognitive radio, where it is required
to meet the following three requirements.
• Fast sp ec tru m sensing. Since it is not possible to transmit and sense
simultaneously at a particular frequency band, sensing has to be interleaved
with data transmission. Hence, the required observation time (or window)
should be as short as possible in order to maximise the overall throughput.
• H igh accuracy. The spectrum sensing device identifies vacant spectrum
bands by detecting presence of primary signal, i.e., by performing a binary
hypothesis test. With noise and fading available in any communication sys
tem, sensing errors are inevitable. For example, false alarm occurs when an
idle channel is detected as busy, and miss detection occurs when an occupied
channel is declared as idle. In the occasion of a false alarm a transmission
opportunity is overlooked, resulting in waste of the spectrum, while miss detec
tion leads to collision with primary users and hence, interference. According
to FCC the performance of spectrum sensing should reach the probability
of detection (PD) of more than 90% (or equivalently the probability of mis-
detection smaller than 10%) whilst the probability of false alarm (PFA) does
not exceed 10%.
• Low com plexity. The computational complexity of the sensing device should
be kept as low as possible in order to reduce the signal processing time, device
energy consumption as well as the infrastructure cost.
In addition, it is becoming increasingly demanding for delivering reliable spectrum
sensing in very low signal-to-noise ratio (SNR) range, such that heavily shadowed
signals can be identified.
1.2. Motivation and Objective
1.2 M otivation and Objective
A lot of research has been carried out in the area of local spectrum sensing, how
ever, it is realised that existing spectrum sensing schemes can hardly meet the
requirements of a fast and accurate spectrum sensing particularly in very low SNR
range, (considering that the target SNR for a reliable spectrum sensing sensitivity
is about —20 dB [5]) without introducing high complexity to the system. This ob
servation motivates us to develop a new local spectrum sensing scheme, which can
significantly improve the state-of- the-art and provides a practical solution, while
showing robustness towards physical impairments which exists in communication
systems, e.g., timing and frequency offsets and noise uncertainty issue.
Furthermore, with increase in the spectrum utilisation, spectrum scarcity increases.
This would call for spectrum sensing techniques that adopt an architecture to si
multaneously search over multiple frequency sub-bands at a time. However, the
literature of sub-band spectrum sensing is rather limited. This observation moti
vated us to develop a local wide-band spectrum sensing spectrum, specifically for
multi-carrier based environments, which can offer reliable performance in low SNR
environments without introducing high complexity to the systems.
1.3 Major Contribution
The major contributions of this thesis are considered to be two-fold.
I. Firstly, a novel spectrum sensing scheme namely, cluster-based differential
energy detection is proposed. It has several distinctive features including low
latency, high accuracy, reasonable computational complexity, as well as robust
ness to very low SNR. Since most of the current and future mobile networks
are multi-carrier based systems, thus the proposed approach has a wide impli
cation for practical applications. The key idea of the proposed scheme is to
exploit the channel frequency diversity inherent in high data-rate communi
cations using the clustered differential ordered energy spectral density (BSD).
Specifically, after the BSD computation, the clustering operation is utilised
to group uncorrelated subcarriers based on the coherence bandwidth to enjoy
a good frequency diversity. The knowledge of coherence bandwidth does not
need to be very accurate. Furthermore, making use of order statistics of the
1.3. Major Contribution
estimated ESD, we further increase the reliability of the sensing algorithm. In
order to exploit the second order moment diversity of the observed signal, a
differential operation is performed on the rank ordered ESD. When the chan
nel is frequency selective and the noise is white, the differential process can
effectively remove the noise floor resulting in ehmination of the noise uncer
tainty impact which is the main factor making energy detection reluctant [9].
At the final stage of the proposed scheme, the differential rank ordered ESD
within different clusters are linearly combined in order to further reduce the
effect of impulse/spike noise. Binary hypothesis testing is then applied on
either the maximum or the extremal quotient (maximum-to-minimum ratio)
depending on the wireless channel characteristics of the sensed environment.
More importantly, the proposed spectrum sensing scheme is designed to allow
robustness in terms of both time and frequency offset. In order to analytically
evaluate the proposed scheme, both PD and PFA are derived for Rayleigh
fading channels. The closed-form expression shows a clear relationship be
tween the sensing performance and the cluster size, i.e., channel coherence
bandwidth, which is an indicator of the diversity gain. Computer simulations
are carried out in order to evaluate the effectiveness of the proposed approach
and to compare the performance of the proposed scheme with state-of-the-
art spectrum sensing schemes where up to 10 dB gain in performance can be
observed.
II. The second contribution, is three-fold; 1) re-defining the objective of the
sub-band level spectrum sensing device to a model estimator, 2) deriving the
optimal model selection estimator for sub-band level spectrum sensing for
fixed and variable number of users along with a sub-optimal solution based
on Bayesian statistical modelling and 3) proposing a practical model selection
estimator with relaxed sample size constraint and limited system knowledge
for sub-band spectrum sensing applications in Orthogonal Frequency-Division
Multiple Access (OFDMA) systems. The proposed technique takes advantage
of the second order moment channel frequency diversity. More interestingly,
it does not require a priori knowledge of noise power and the propagation
channel gain, and is designed in such a way to show robustness towards energy
leakage. The proposed model selection based sub-band level spectrum sensing
approach is analytically evaluated through PFA and PD along with closed
form expression for the threshold value. Furthermore, computer simulations
are carried out in order to evaluate the effectiveness of the proposed scheme.
1.4. Thesis Organisation
1.4 Thesis Organisation
The rest of thesis is organised as follows, in chapter 2 the existing approaches
for local spectrum sensing are reviewed, a novel cluster-based differential energy
detection for spectrum sensing in multi-carrier systems is proposed in Chapter 3.
Furthermore, in Chapter 4 a Bayesian model based approach for joint sub-band
level spectrum sensing is presented. Finally, conclusions are drawn in Chapter 5
along with the set of possible future work.
1.5 Publication List
1.5.1 Journal Publications
P. Cheraghi, Y. Ma, and R. Tafazolli, and Z. Lu, “Cluster-based Differential Energy
Detection for Sensing Multi-carrier Sources in Very Low SNR”, IEEE Transaction
on Signal Processing, To appear, December, 2012.
P. Cheraghi, Y. Ma, and R. Tafazolli, and Z. Lu, “A Bayesian Model Based Ap
proach for Joint Sub-Band Level Spectrum Sensing”, IEEE Transaction on Signal
Processing, (Submitted, July 2012)
Z. Lu, Y. Ma, P. Cheraghi, and R. Tafazolli, Novel Pilot Assisted Spectrum Sensing
for OFDM Systems by Exploiting Statistical Difference between Subcarriers,” IEEE
Transaction on Communications, (First revision submitted, September, 2012)
P. Cheraghi, Y. Ma, R. Tafazolli and Z. Lu, “Spectrum Sensing in OFDM environ
ments: Advances and Challenges” , IEEE Journal on Selected Areas in Communi
cations, (To be submitted, October 2012)
1.5.2 Conference Proceedings
P. Cheraghi, Y. Ma, and R. Tafazolli, “A Novel Blind Spectrum Sensing Approach
for Cognitive Radios,” PCNET 2010, 21-22, Jun., 2010
P. Cheraghi, Yi Ma, and Tafazolli, R. , “Frequency-Domain Differential Energy
Detection Based on Extreme Statistics for OFDM Source Sensing,” 2011 IEEE
73rd Vehicular Technology Conference (VTC Spring), pp.1-5, 15-18 May 2011
1.5. Publication List
P. Cheraghi, Y. Ma, Z. Lu, and R. Tafazolli, “A Novel Low Complexity Differential
Energy Detection For Sensing OFDM Sources In Low SNR Environment,” 2011
IEEE GLOBECOM Workshops (GO Wkshps), pp. 378-382, 5-9 Dec. 2011
Z. Lu, P. Cheraghi, Y. Ma, and R. Tafazolli, “Extreme Statistics Based Spectrum
Sensing for OFDM Systems by Exploiting Frequency-Domain Pilot Polarity,” 2011
IEEE GLOBEGOM, 5-9 Dec. 2011
1.5.3 Patents
International Patent No: PCT/CB2012/050764 , “Spectrum Sensing of OFDM”,
filed 4 April 2012.
Chapter 2
State-of-the-art Spectrum Sensing
Techniques
The basic idea behind development of opportunistic spectrum access technologies
such as cognitive radio is to increase the spectral utilisation. This goal can be
achieved by identifying and utilising the spectrum holes; given the, the conventional
definition of spectrum hole: “a band of frequencies that are not being used by the
licensed user of that band at a particular time in a particular geographical area” [4].
Hence, based on this definition, identifying spectrum holes requires exploitation of
three dimensions of frequency, time and space by the spectrum sensing device. In
other words, the main task of any spectrum sensing device is to determine if the
frequency band of interest is occupied by the licensed user during a time slot within
a certain geographical area [10].
In what follows in this chapter, firstly, we will consider the original problem of
local narrow-band spectrum sensing and provide a comprehensive study of existing
solutions and secondly, the problem of local sub-band sensing is considered along
with the state-of-the-art sub-band spectrum sensing algorithms.
2.1 Problem Formulation for Narrow-Band Spec
trum Sensing
The general problem of spectrum sensing can be modelled as the binary hypothesis
testing with hypothesizes: when the frequency band of interest is vacant and
2.1. Problem Formulation for Nairow-Band Spectrum Sensing 8
no other users is utilising this frequency band; and 77i, when the frequency band
of interest is occupied by other users and not available for opportunistic usage.
Therefore, this problem mathematically reads as
(2.1)y s{t) + v{t), Hi
where y{t) denotes the received signal at the spectrum sensing device, v{t) is the
Additive White Gaussian Noise (AWGN) with zero mean, and s{t) represents the
signal transmitted by the existing active users. Hence, based on (2.1) spectrum sens
ing techniques can be evaluated through two classical metrics, namely probability
of detection (PD) and probability of the false alarm (PFA).
Mathematically, the probability of false alarm is defined by [11]
P F A A P r ( . f > A | % ) , (2.2)
where ^ denotes the test statistics and A is the detection threshold. As it can be
observed from (2.2), PFA refiects the probability of an event where the spectrum
sensing device reports an alarm when the signal is actually not being transmitted.
While, probability of detection is defined by
PD A P r ( . ^ > A I % ) . (2.3)
From (2.3) it can be concluded that PD reflects the probability of an event where
the spectrum sensing device reports an alarm when the signal is indeed there.
Spectrum sensing problem explained in (2.1) has been quite well investigated in the
literature, for which there are many approaches reported. In general the existing
spectrum sensing approaches can be divided into three main categories:
• Exploiting the energy difference
• Exploiting the stationarity difference
• Exploiting the distribution difference.
In the following sections each of the above categories is explained and compared in
detail.
2.2. Exploiting Energy Difference
2.2 Exploiting Energy Difference
The spectrum sensing approaches which fall into this category make a decision based
on the estimated energy of the received signal at the spectrum sensing device. The
most well-known spectrum sensing approach developed under this category is the
energy detection [11].
2.2.1 Energy Detection
Average DecisionMakingA/DFilter
Figure 2.1: Flow chart of conventional energy detection.
y(n)FFTA/D Average Decision
Making
Figure 2.2: Flow chart of frequency-domain energy detection.
Energy detection (radiometer) is the simplest spectrum sensing approach introduced
in the literature [11]- [15]. Due to its low computational complexity it is also the
most common technique used in practice. This approach can be thought of as a
blind spectrum sensing approach since it does not require a priori knowledge of
the signal being detected. Therefore, it is robust to variation of the signal being
detected and thus it is known as the optimal detector in the case where we assume
absolutely no deterministic knowledge about the signals being detected. Figure 2.1
and Figure 2.2 show block diagrams of the energy detection algorithm. As it can be
observed this technique can be implemented in either frequency or time domain. It
is worth mentioning that the conventional energy detection is implemented in time
domain, while the frequency domain version was later introduced for sub-band [16]
purposes. Frequency domain energy detection is usually used to simultaneously
detect the vacancy of several sub-bands, i.e. filter bank based energy detection [16].
The test statistics ^ ed for time domain energy detection is given by
(2.4)n = 0
2.2. Exploiting Energy Difference 10
where y(n) and N represents the sampled received signal and the observation length
respectively. While the test statistics in frequency domain energy detection is also
calculated as (2.4), with a minor difference that it is a function of ÿ = ^ n x n Y
given that ^ n x N denotes an N x N Discrete Fourier Transform (DFT) matrix and
y = [?/(0),2/(1), y{N - 1)]^.
Using central limit theorem [17] it can be observed that the alternative hypothesis
testing in (2.1) for energy detection can be expressed as [9]
(“ )
where and denote the noise and signal variance respectively. As it can be
observed from (2.5) the energy difference which exists between the two hypotheses
can be used as a metric for performing spectrum sensing. Thus, the availability of
the frequency band of interest using energy detection approach can be determined
using
^ E D >^ED (2 .6 )
where X e d denotes the energy detection threshold. It is clear from (2.5) and also
from the derivations in [11] and [9] that the threshold value is directly proportional
to the noise power. Thus, it can be concluded that the energy detection approach
requires accurate knowledge of the noise power, specifically in low SNR values, in
order to deliver a reliable performance.
However, noise power estimation error is unavoidable in practical systems, hence,
this will give rise to a phenomenon called noise uncertainty [11] [9].
2.2.2 Energy Detection Under Noise Uncertainty
In most communication systems noise is an aggregation of various independent
sources, i.e. thermal noise, interference due to nearby unintended emissions, etc.
Thanks to the central limit theorem [17], one can assume that noise at the receiver
is a Gaussian random variable. We should bear in mind that the error due to this
assumption will tend to zero as where N is the number of independent random
variables being summed up. In a practical scenario N is usually moderate therefore
this error can not be neglected, especially in low SNR environments. For some
2.2. Exploiting Energy Difference 11
Test Statistics (Estimated Energy)
III
Target Signal
. Noise power ^ uncertainty zone
Figure 2.3: The SNR wall phenomenon.
constant K the error due to this assumption can be modelled as [17]:
K\F v ix ) - A T { x ) \< - ^ (2,7)
where Fv{x) denotes the actual noise distribution and M{x) denotes a zero mean
Gaussian distribution. Nevertheless, most detectors operate under the assumption
of the received noise being Gaussian. The other main factors causing noise uncer
tainty is the temperature variations at the receiver which leads to inaccurate noise
power measurements. Hence, it can be concluded that if the SNR at the spectrum
sensing device is sufficiently low, there would be enough uncertainty in the noise
to render the energy detection useless. Consider the case with noise uncertainty
factor C/, since the energy detector only sees the the energy, distribution for the
noise uncertainty can be summarised in a single interval = [(l/f/)(j„2 , U a^.
It is further proved in [9] that as the observation length is increased, i.e., N oo,
the minimum operating SNR in which the energy detection can operate desirably
will converge to S N R I (U — (^))- This introduces a phenomenon called “«SiVR
w a ir . SNR wall of a spectrum sensing device is defined as the maximal SNR such
that for any SNR < SNRwaii detection is impossible for that particular detector.
The SNR wall for energy detector can be expressed as:
SNRwaii =C /2-1
U(2.8)
In the last four decades since the publication of [11], many solutions have been
2.3. Exploiting Stationarity Difference 12
developed to make energy detection more robust in terms of SNR wall (e.g. [18]-
[20]), yet the noise uncertainty problem in spectrum sensing approaches based on
the energy difference still exists. Hence, small noise power estimation error can
result in significant performance loss [21]. Consequently the noise power has to be
estimated dynamically. This is done by separating the noise and signal subspaces
using multiple signal classification algorithm [22]. Noise variance estimated by using
the value of the smallest eigenvalue of the incoming signals autocorrelation. In
[23] an iterative algorithm is proposed to find the near optimum threshold value.
The performance of energy detection technique over various fading channel models
has been investigated in [14], where it is shown that fading channels can have
a detrimental effect on the performance of the energy detection based spectrum
sensing.
2.3 Exploiting Stationarity Difference
Stationarity is defined as a quality of a process in which the statistical parameters
of the process do not change with time [24]. The spectrum sensing approaches
which fall into this category exploit the various stationarity difference which exists
between the noise and signal. In what follows the state-of-the-art spectrum sensing
approaches which fall into this category are explained in detail.
2.3.1 Second Order Cyclostationarity Detection
y(n)A/D Decision
MakingAuto
correlation FFT
Figure 2.4: Flow chart of second order cyclostaionary based detection technique.
The initial works of spectrum sensing through stationarity difference can be traced
back to work of Dandawate et al. in [25], where second order cyclostionarity is
employed. Cyclostationarity detection is based on exploiting the cyclostationarity
feature of the received signal [26]- [27]. This feature is caused by periodicity in the
statistics of the transmitted signal which could be a result of modulation, coding
2.3. Exploiting Stationarity Difference 13
or intentionally to assist spectrum sensing [28]- [30]. Process y{n) is considered sub
sense cyclostationary process if
Py = E [y{n)] = E [y{n -f V)] (2.9)
and
Ry (rj) = E [y{n)y* (n + 7])]
- E [y{n + V)y* {n + V + rj)] (2.10)
Vn, rj E Z . The smallest value of V for which (2.9) and (2.10) hold is called the
period. Being periodic, Ry{n) follows Fourier Series expansions over cyclic frequen
cies with the set of cycles A := [>c = A; = 0 ,1 , . . . , "P — 1]. Hence, the Fourier
coefficients also called cyclic auto correlations are related to Ry{n) using
1(^) Gxp { - p n x n ) . (2.11)
n = 0
As it can be observed from (2.11) the cyclic autocorrelation function at a given
cyclic frequency determines the correlation between spectral components of the sig
nal separated in frequency by an amount of hence, given sufficient observation
length the cyclic autocorrelation function of cyclostationary signals is nonzero only
for set of cycles which fall in the set A. Thus, one can determine the vacancy of
the frequency band of interest by analysing the cyclic autocorrelation function of
the received signal [25], i.e. using the second order cyclostationary feature. More
interestingly, cyclostionarity based detection can be employed to differentiate differ
ent types of signals [31]. Figure 2.4 illustrates the block diagram of a second order
cyclostationarity based detection technique.
In the above procedures it has been assumed that the cyclic frequency is known
at the receiver. However, this assumption may not be reasonable when the spec
trum sensing device is required to perform sensing in a sub spectrum band. Since,
the sensing device may not have knowledge of the period of all the users operating
and storing these information covering the whole geographical area in which the
sensing device (specifically for non stationary users) requires connection to an accu
rate location aided database. Thus, an exhaustive search is required to determine
the operating cyclic frequencies. This will increase the complexity of the detector
significantly and furthermore the detector will loss the ability of distinguishing be
2.3. Exploiting Stationarity Difference 14
tween the signal and the interference which also have cyclic characteristics. It is
shown in [29]- [25] that the cyclostationarity based scheme can trade latency with
high sensing reliability. Furthermore, It is less sensitive to the noise uncertainty
provided the knowledge of signals cyclic frequency [32].
2.3.2 Covariance Based Detection
yi(t)
DecisionMaking
Qmax
Qmin
Figure 2.5: Flow chart of covariance based detection.
To overcome the requirement of cyclic frequency information in cyclostationary
based detection while not suffering from the noise uncertainty problem which exists
with energy detection, the covariance based detection was introduced [33]. This
spectrum sensing method, as the name implies, is based on the estimated covari
ance matrix of the received signal, and utilises the correlation which exists in the
transmitted signal to determine the vacancy/occupancy of the frequency band of
interest. Hence, it is proved to be very effective when the transmitted signal is
highly correlated [33]- [34].
The correlation of the received signal samples can be due to many factors, e.g.,
modulation, multi-path fading, multiple receivers or can be intentionally introduced
by oversampling at the spectrum sensing device [35], while the noise samples are
independent. Moreover, since the covariance matrix of noise is determined by the
receiving filter at the receiver its structure is known to the spectrum sensing device,
allowing us to differentiate the two hypothesis in (2.1) [36].
Consider L to be the number of consecutive samples used for estimation the covari
ance matrix, i.e..
y{n) = [y{n) y { n - l ) . . . y { n - L + l)y (2.12)
2.3. Exploiting Stationarity Difference 15
hence, the estimated covariance matrices can be expressed as
\ \ ^ C y = C s + a % (2.13)
where
C , =E[s(n)s*(n)] (2.14)
Cy =E[y(n)y*(n)] , (2.15)
1l denotes an L x L Identity matrix and s(n) = [s{n) s{n — 1) . . . s(n — L + 1)]^.
Given that the noise samples are uncorrelated based on (2.13), it can be concluded
that considering the 'Ho scenario in (2.1), if the signal s(t) is not present, i.e. when
Cg = 0, the off-diagonal elements of Cy are all zero. On the other hand, if the
signal s(n) is present, i.e. 'Hi hypothesis in (2.1), Cy will no longer be a diagonal
matrix due to the correlation between the s(n) samples, resulting in off diagonal
elements. Hence, the vacancy of the frequency band of interest can be determined
based on the sum value of the off-diagonal elements. Based on [33], an effective test
statistic for this purpose would be ^ where
Ï Ê É (2.16)n = 0 m = 0
1Î2 = T E K " l' (2.17)
n = 0
given that Cm,n denote the element of the matrix Cy at the mth row and nth
column.
Considering the 'Ho scenario, = 1 while given 'Hi hypothesis 3 cd > 1 •
However, this is based on the assumption that L ^ oo, which is not a practical
assumption, hence in order to make a reliable decision 3 cd should be compared
to a threshold value which is function of observation length and the required PFA.
Details of the threshold setting can be found in [33]. It should be noted that the
performance of the covariance based detection is highly dependent on the correlation
of the received signal hence, in the extreme case where the received sample are
Independent and Identically Distributed (i.i.d.) this spectrum sensing approach
will fail.
2.3. Exploiting Stationarity Difference 16
2.3.2.1 Eigenvalue Based D etection
yi(t)
p ( t )
yM(t)
Iyi(n)
yzM
yM(n) u
Eigenvalue Qmax DecisionDecomposition Qm in Making
Figure 2.6: Block diagram of Eigenvalue based detection technique.
Following the development of the covariance based spectrum sensing approach the
eigenvalue based detection was introduced in [35]- [38]. The eigenvalue based detec
tion scheme exploits orthogonality between the signal subspace and noise subspace
using covariance matrix, i.e., second order stationarity features, to offer highly reli
able spectrum sensing [35].
Hence, based on this approach the vacancy of the frequency band of interest is
determined based on the fluctuation of the eigenvalues [24] of the covariance ma
trix, and hence many test statistics have been proposed to efficiently utilise this
fluctuation [35]- [38].
Amongst the well known test statistics for eigenvalue based detection is the ratio of
the maximum to minimum eigenvalues [35]. Let Pmin and Pmax denote the minimum
and the maximum eigenvalues of the covariance matrix Cy, respectively, hence, the
availability of the spectrum can be determined using
^E v = Aay. (2.18)
In [35], the asymptotic statistical characteristics of gmax and gmin under Hq have
been investigated and furthermore, closed from expression of the decision threshold
X e v for a given probability of false alarm has been provided.
An other popular test statistic used in eigenvalue based detection is the ratio of the
maximum eigenvalue to the average eigenvalue, i.e, ^ e v = where
(2.19)1=0
2.3. Exploiting Stationarity Difference 17
The asymptotic threshold value of the above test statistic is provided in [35]. How
ever, like covariance based detection, the eigenvalue based detection will fail if and
only if C5 = (TsIl i.e., the received signal samples are i.i.d. . However, the correla
tion between the received signal can be forced with employment of multiple receive
antennas or oversampling.
2.3.3 Matched Filtering
Matched filtering is known to be the optimum method for detecting signals when the
transmitted signals air interface is completely known to the sensing device [39]. The
main advantage of matched filtering is low latency and computational complexity.
Match filtering requires a very short time to achieve the desired probability of false
alarm or probability of detection as compared to all the existing detection methods
while introducing a linear complexity even in a very low SNR environment since
it maximise the received SNR at the sensing device [39]. On the other hand the
shortcoming of the match filtering technique is that the spectrum sensing device
needs to demodulate the received signal prior to determine the vacancy of the
frequency band of interest. Hence it will introduce to main concerns 1) requiring
a perfect knowledge of the transmitted signals signalling feature, 2) security issue,
since this allows the spectrum sensing device to have access to the transmitted
message.
Since the spectrum sensing device should be able to detect all the possible signals
transmitted within the bandwidth of interest, the implementation complexity of
such detection technique is impractically large, due to perfect knowledge of all the
available signalling information of users [40]. One of the other disadvantages of
match filtering is large power consumption as all the possible receiver algorithms
needs to be executed for spectrum availability decision making.
2.3.3.1 Pilot Based M atch Filtering D etection
In practical communication systems, pilots are usually transmitted periodically for
time or frequency synchronisation applications, channel estimation, etc. These pi
lots, if known to the spectrum sensing device, can be utilised for coherent detection
2.4. Exploiting The Distribution Difference 18
y(n)CorrelationA /D Decision
Making
Figure 2.7: Block diagram of the pilot based detection technique.
of the transmitted signal with the aid of match filtering. Therefore, it works even un
der a very low SNR region. Furthermore, it has lower complexity and latency than
statistics based cyclostationary and covariance based detection while overcoming
the noise uncertainty problem. Moreover, it does not require demodulation of the
transmitted signal as the conventional match filtering since orthogonal to the data
and can be considered independently. Therefore, pilot based coherent detection is
always one of the preferred spectrum sensing schemes in practice.
In the scenarios where the pilot structure is known to the spectrum sensing device
the optimum detection technique would be match filtering. Therefore, the test
metric can be expressed as:
S^PM = ~^{y{n)sl{n)} (2 .20)
where Sp{ t ) denotes the known pilot signal. Hence, a decision can be made using,
^PM ^PMi where \p m denotes the threshold value to satisfy the required
probability of false alarm.
In (2.20) perfect synchronisation between the sensing device and the transmitter is
assumed, while this condition is not feasible in practice. Hence, the sensing device
has to perform an exhaustive search to find the timing offset which maximises the
^PM value. Recently, various robust pilot-based coherent detection schemes have
been proposed for spectrum sensing applications [41]- [43] .
2.4 Exploiting The Distribution Difference
Given that in almost all communication system models, noise is assumed to be ad
ditive white and Gaussian, one can determine the vacancy of a particular frequency
band by observing the difference of the received signals distribution and that of the
AWGN. Based on this feature a number of well known spectrum sensing algorithms
have been proposed, which are explained in this section.
2.4. Exploiting The Distribution Difference 19
2.4.1 Entropy Based Detection
y(n)Partition Decision
MakingFFTA/D EntropyEstimation
Figure 2.8: Flow chart of entropy based detection.
In information theory, entropy is a measure of the uncertainty associated with a
discrete random variable. The term by itself usually refers to the Shannon entropy,
which quantifies the information conveyed in a message [44]. Recently, entropy based
detection approaches have been employed for spectrum sensing applications [45] [46].
Entropy based spectrum sensing can be thought of an approach which exploits the
distribution difference in order to determine the vacancy of the frequency band of
interest.
In order to allow robustness to noise uncertainty issue, the entropy based detection
makes a decision based on the estimated entropy of the measured signal in the
frequency domain with the probabihty space partitioned into fixed dimensions. This
is due the fact that the entropy of the received signal in the time domain is related to
the signal power and is sensitive to noise uncertainty [45]. Hence, the test statistic
for this spectrum sensing approach can be expressed as [46]
L-l= Xe b ,
i=0(2.21)
where L denotes the dimension of probability space, N is the number of Discrete
Fourier Transform (DFT) points, k{ is the total number of occurrences at the
probability state, and Xe b is the threshold value used for decision making. As
suming that the estimated noise entropy follows a Gaussian distribution, the value
of Xe b can be easily calculated based on the desired PFA and value of L [46].
However, the entropy based detection will fail to deliver accurate results, when the
transmitted signals also follow a Gaussian distribution and since the convergence to
normality could be extremely slow, this approach will require relatively high obser
vation length. However, this would is in contradiction to achieving channel capacity,
which benefit from Gaussian noise like transmit signals.
2.4. Exploiting The Distribution Difference 20
2.4.2 Kurtosis Based Detection
y MA/D Decision
MakingK urtosis
Estim ation
Figure 2.9: Flow chart of kurtosis based detection.
In statistics, kurtosis is a descriptor of the shape of a probability distribution, i.e.,
it is a measure of the “peakedness” of the probability. Hence, the kurtosis based
detection was introduced to exploit the non-Gaussianity of communication signals
in order to determine the availability of the frequency band of interest [47]- [48]. For
example for randomly occurring signals that produce non Gaussian distributions,
the kurtosis estimate can be less than 3 or it can have a value much greater.
This scheme features excellent accuracy at the price of large latency due to higher-
order statistics. A critical point is that the sensing performance degrades signifi
cantly when signals are approximately Gaussian. Kurtosis is defined by the ratio
of the expected value of the fourth-order central moment and the square of the
expected value of the second-order central moment. Hence, the test statistic of the
kurtosis based detection can be formulated as
E \{y{n) - f l y ) ] ^KB = — -------------- ^
E [(y(n) - fly) J(2 .22)
This scheme features excellent accuracy at the price of high computational complex
ity due to higher-order statistics. Furthermore, the convergence to normality could
be extremely slow, and the sample estimate of the kurtosis can deviate substantially
from its true value even with a large number of observations. Thus, for moderate
sample sizes which is a prerequisite for any spectrum sensing device, the kurtosis
test cannot be expected to be accurate.
2.5. Summary of The Narrow-Band Spectrum Sensing Approaches 21
2.5 Summary of The Narrow-Band Spectrum Sens
ing Approaches
In the above discussion, we have introduced various state-of-the-art narrow-band
spectrum sensing techniques. As explained, various spectrum sensing techniques
have different advantages/ disadvantages and hence are applicable in different sens
ing scenarios. In general, the existing spectrum sensing approaches can be divided
into three main categories:
• E xplo iting energy difference. The most well-known spectrum sensing
approach developed under this category is the energy detection [11]. The
energy detection is recognized as a blind sensing scheme with advantages such
as low complexity and low latency. However, it is very sensitive to the noise
uncertainty such that its performance is limited by the SNR wall [9]. In the last
four decades since the publication of [11], many solutions have been developed
to make energy detection more robust in terms of SNR wall (e.g. [18]- [20]),
yet the noise uncertainty problem in spectrum sensing approaches based on
the energy difference still exists.
• Explo iting s ta tio n a rity difference. The initial works of spectrum sensing
through stationarity difference can be traced back to work of Dandawate et
al. in [25], where second order cyclostionarity is employed. The cyclosta
tionarity based scheme can trade latency with high sensing reliability. It is
less sensitive to the noise uncertainty provided the knowledge of signals cyclic
frequency [32]. To overcome the requirement of cyclic frequency in cyclosta
tionary based detection while not suffering from the noise uncertainty problem
which exists with energy detection, the covariance based detection was intro
duced [33]. This spectrum sensing method, utilises the correlation which exists
in the transmitted signal to determine the vacancy/occupancy of the frequency
band of interest. However, the performance of this approach degrades dramat
ically as the correlation of the transmitted signal decreases. Matched filtering
is known to be the optimum method for detecting signals when the transmit
ted signals air interface is completely known to the sensing device [39]. The
main advantage of matched filtering is low latency and computational com
plexity. However, this approach requires perfect synchronisations between the
transmitter and the spectrum sensing device. Matched-filtering pilot based
2.5. Summary of The Narrow-Band Spectrum Sensing Approaches 22
Approach Air-InterfaceInformation Synchronisation Computational
Complexity Latency
Energy Detection X X Low Low
Cyclostationarity Detection / / High High
Covariance Based Detection X X Medium Low
Eigenvalue Based Detection X X High Medium
Matched Filtering / / Low Low
Pilot Based Matched Filtering / / Low Low
Entropy Based Detection / X Medium Medium
Kurtosis Based Detection / X High High
Table 2.1: Summary of the state-of-the-art local narrow-band spectrum sensing approaches.
detection, given the knowledge of pilot symbols and reasonably good timing
and frequency synchronizations, exploits the cyclostationary property of the
pilot symbols, to deliver fast and reliable sensing. The eigenvalue-based de
tection scheme exploits orthogonality between the signal subspace and noise
subspace using second order stationarity features to offer highly reliable spec
trum sensing [35]. However, it often needs the support of multiple antennas,
and the subspace decomposition costs cubic complexity.
• Explo iting th e d istrib u tio n difference. Given that in almost all commu
nication system models, noise is assumed to be additive white and Gaussian,
one can determine the vacancy of a particular frequency band by observing
the difference of the received signals distribution and that of the AWGN. An
example of such approaches would be the kurtosis-type scheme, which exploits
the non-Gaussianity of communication signals [47]- [48]. This scheme features
excellent accuracy at the price of large latency due to higher-order statistics.
A critical point is that the sensing performance degrades significantly when
signals are approximately Gaussian. Entropy based spectrum sensing can be
thought of an approach which also benefits from this property [45], where
the probability space is partitioned into fixed dimensions and the Shannon
entropy is employed as the information measure of the received signal as the
test statistic.
2.6. Sub-Band Level Spectrum Spectrum Sensing 23
2.6 Sub-Band Level Spectrum Spectrum Sensing
With increase in the spectrum utilisation, spectrum scarcity increases. This would
call for spectrum sensing techniques that adopt an architecture to simultaneously
search over multiple frequency sub-bands at a time, while meeting the mandatory
requirements of spectrum sensing, i.e.,1) low latency, 2) high reliability and 3) low
complexity.
However, the literature of sub-band spectrum sensing is rather limited at this time.
In this section we will provide a system model for sub-band level spectrum sensing
and further explain the state-of-the-art sub-band level spectrum sensing techniques
in detail.
2.6.1 Sub-Band Level Spectrum Sensing System Model
Consider a communication system operating over a sub-band channel that is di
vided into K non-overlapping sub-bands, e.g., multi-carrier systems. However, in
a particular geographical region within a certain time frame only I number of the
sub-bands are utilised by the users, where I < K . Thus, {K — I) sub-bands are
available for opportunistic access. The essential task of the spectrum sensing device
is to determine the availability of these {K — I) sub-bands.
Let y{n) denote the received sub-band signal at the spectrum sensing device. Hence:
K - l
2/W = ^ AfcSfc(n) 4- v{n) (2.23)fe=o
where Aq is the indicator function which denotes the presence of the transmitted
signal q. The opportunistic user needs to determine which of the spectrum bands
are unoccupied, in order to utilise them efficiently. Based on (2.23) a number of
solutions were proposed in the literature which are fully explained in the following
sections.
2.6.2 Filter-Bank Based Spectrum Sensing
The Filter-bank architecture allows sub-band sensing with the aid of multiple narrow
band, band-pass filters [16] [49]. Filter banks are often implemented based on a
2.6. Sub-Band Level Spectrum Spectrum Sensing 24
prototype Iter. The prototype filter is a lowpass Iter that is also used to realise the
first sub-band of the filter bank. Other bands are realised through repetition of the
prototype filter. Hence, all the N sub-bands of interest share the same structure.
The implementation of an spectral estimator that uses a filter bank for signal anal
ysis is as follows: 1) the input process is passed through a bank of filters and 2) the
output power of each filter is measured as an estimate of the spectral power over the
associated sub-bands and finally the vacancy of each sub-band is determined based
on the estimated power of that particular sub-band. However, like conventional
energy detection this approach will face the noise uncertainty problem.
It is shown in [49] that the filter bank based spectrum sensing performs significantly
better in filter bank-based multi-carrier communication systems, since the same
filters can be utilised for sensing purposes. Hence, in such systems, channel sensing
is done at virtually no cost. This is only possible given that all the users within
the geographical area of interest share the same air-interface and furthermore the
opportunistic user also employs the same air interface. However, in a more general
case where users may have different signalling format the filter-bank approach will
result in increased number of components and energy consumption.
2.6.3 Joint Multi-Band Detection
In order to improve the performance of the filter bank detection the joint multi
band detection, was proposed in [50]. This approach jointly optimises a bank of
multiple narrow band detectors to improve the aggregate opportunistic throughput
of the opportunistic users while limiting the interference to the existing users. In
particular, the joint multi-band detection reformulates the original problem of sub
band spectrum sensing into a class of optimisation problems, where the objective
is to maximise the aggregate opportunistic throughput in an interference-limited
network given the opportunistic rate and interference penalty on each sub-band are
known to the spectrum sensing device. Hence, the optimisation problem can be
summarised by [50]
2.6. Sub-Band Level Spectrum Spectrum Sensing 25
[1 — PFA (7 )]
K - l
s.t. : ^ 2 ^ [1 ~ FD j(7i)] < €j , j = 1 ,2 ,. . . Ji=o (2.24;
PFA (7 ) < a
[ 1 - P D (7 )]< /S
where r = [ r i ,r2, . . . ,tk]'^ is a vector with the throughput achievable over all K
sub-bands, 7 = [71,72, - - - , 7k]^ is the vector denoting the threshold value for all
sub-bands, 1 —a = [ l - a i , l — o:2,---,l — ock]^ and f3 = [Pi, /32, . . . , Pk ]" are the
minimum limit for opportunistic spectral utilisation required from the spectrum
sensing device and the upper limit for the interference introduced by the oppor
tunistic users, respectively.
Hence, the threshold setting in this approach is in a such a way to firstly assure
that the sub-band with a higher opportunistic rate has a higher threshold . In
other words reduce PFA for the corresponding sub-band to ensure best possible use
by the opportunistic users. Secondly, the higher priority sub-band, i.e., sub-bands
carrying important messages, have a lower threshold resulting in smaller PD in order
to prevent opportunistic users interference. Finally, a little compromise on the sub
bands carrying less important information which might boost the opportunistic rate
considerably. Thus, in the determination of the optimal threshold for each sub-band,
it is necessary to balance the channel conditions, the opportunistic throughput,
and the relative priority of each sub-band. It has been shown in [50] that the
joint multi-band detection can improve the performance of the filter-bank spectrum
sensing significantly, and that the performance of this approach can be improved
considerably by further exploiting the spatial diversity, i.e., cooperation between
the spectrum sensing devices.
However, this technique requires the knowledge of noise power and the squared
values of the channel frequency responses, which makes this approach only practical
in fixed wireless networks, i.e., TV broadcast bands.
2.6.4 Sequential Multi-Band Detection
In, [51] a sequential detection scheme has been developed for multi-band spectrum
sensing. This approach employs a bank of sequential probability ratio tests [52],
i.e., one per sub-band. The sequential probability ratio test has a very simple
2.6. Sub-Band Level Spectrum Spectrum Sensing 26
structure where the likelihood ratio of the observed samples is tested against two
thresholds. The sequential probability is known to minimise the average sample
number amongst all detectors given the PD and PFA requirements of the system.
Hence, this algorithm can be particularly useful in delay sensitive applications. The
hypothesis testing for this approach for the sub-band can be expressed as
logB^ < < log take the next sample
(2.25)
where the test statistic ^SM is the likelihood ratio estimated from the received
signal and the threshold values are related to the false alarm probability and the
miss detection probability, i.e., [52]
PD*PFA^
1 - P D '1 - PFAk' (2.26)
However, the key challenge associated with this detector is that the parallel sequen
tial probability tests do not yield the same sample sizes. This is due to the fact
that the observation length is variable which depends on the random received sig
nal. Thus, the overall sensing delay will be considered as the largest detection delay
among those of the parallel detectors, until the set of bands that can support the
requested rate is discovered.
2.6.5 Wavelet Based Detection
y(n) WaveletTransfrom
PSDEstimation
DecisionMaking
EdgeDetectionA/D
Figure 2.10: Block diagram of the wavelet based detection technique.
The wavelet based spectrum sensing is able to perform sub-band sensing with the
aid of edge detection [53] [54]. Assuming that the power spectral characteristic is
smooth within each sub-band but exhibits a discontinuous change between adjacent
sub-bands, wavelet based detection has been proposed to identify and locate the
spectrum holes by exploiting the irregularities within the estimated Power Spectral
2.6. Sub-Band Level Spectrum Spectrum Sensing 27
Density (PSD) [24] with the aid of the wavelet transform, an attractive mathemati
cal tool for analysing singularities and irregular structures of signals. Wavelet based
detection has proved useful for fast coarse spectrum sensing based on a number of
non-stationary samples, by making use of the signals non-stationarity features.
The wavelet based detection has been developed under four main assumptions: 1)
The total bandwidth for detection is known to the spectrum sensing device , 2)
The number of licensed users are unknown to the spectrum sensing device, 3) The
PSD of all occupied sub-band is smooth and almost flat, 4) The noise is AWGN, i.e.
noise process has a flat PSD within the whole observed bandwidth.
Hence, once the region of support is determined, wavelet-based approach will firstly
estimate the PSD of the received signal and determine the number of sub-bands and
the corresponding frequency boundaries. Later, the PSD for each sub-band will be
employed to determine the vacancy of the estimated sub-bands. As a result, the
wavelet based detection is also known as the wavelet based edge detection. Hence,
the availability of the sub-band can be determined using the
^WB — y - V - t d / ^WB (2.27)Jk — Jk-1 '//k-l
where fk — f k - i denotes the estimated frequency boundaries of the k*^ sub-band
obtained using the wavelet transform [54] [53]. Based on the above, one of the
main advantages of the wavelet based detection is that it does not require any prior
knowledge about the signals features. However, the most important limitation
of this spectrum sensing approach is determining the correct smoothing function
(mother wavelet) for the wavelet transformation. Even though some common fea
tures are shared by most mother functions, some can perform better than others
in a given environment. Hence, in order to obtain the best possible results in the
wavelet based detection, the specific wavelet family should be designed based on
the characteristics of the transmitted signal.
2.6.6 W igner-Ville Based Detection
The Wigner-Ville based spectrum sensing [55] derives a greyscale image of the time-
frequency description of the received signal through the Wigner-Ville transform,
and similar to wavelet based detection with the aid edge detection is able to detect
occupied frequency bands.
2.7. Summary of The Sub-Band Spectrum Sensing Approaches 28
With the aid of the Wigner-Ville transform, it is possible to show the spectral
components of a signal with respect to the time variable and therefore have a bi-
dimensional description of the perceived signal [56]. The resulting image from the
Wigner-Ville based detection shows the spectrum occupancy in both time and fre
quency, marking the occupied zones with higher brightness. Hence, such zones are
to be avoided by the opportunistic user, who, thanks to an edge detection, is able
to detect the vacant sub-bands.
Such two dimensional strategies such as Wigner-Ville and wavelet based detection,
tend to improve the performance of the spectrum sensing device with respect to sin
gle dimensional approaches due to the phenomenon known as “uncertainty relation
ship” which describes the trade off between the spectral and temporal resolution.
At the final stage, the measured energy level is employed as the slot availability
criterion, entailing that slots are considered occupied even when they present high-
energy even in a narrow spectral and time components. However, these approaches
may suffer from noise uncertainty problem, due to use of energy detection.
2.7 Summary of The Sub-Band Spectrum Sensing
Approaches
As explained in this section, different spectrum sensing scenarios demand for differ
ent sub-band spectrum sensing approaches based on their requirements. Amongst
the existing practical solutions, there are filter-bank [49] and wavelet [54] based
spectrum sensing techniques. The Filter-bank architecture allows sub-band sens
ing with the aid of multiple narrow bands, which results in increased number of
components and energy consumption. However, the filter-bank approach is one
of the preferred solutions when the spectrum sensing device does not have any
a-priori knowledge about the signalling information. Wavelet based spectrum sens
ing exploits the multi-resolution features of the wavelet transform to estimate the
power spectral density. With the aid of edge detection, the spectrum band of in
terest is divided into a number of sub-bands. This technique is particularly useful,
when there are limited number of non-stationary samples. Fine spectrum sensing
is further required, in order to determine the vacancy of specified frequency sub
bands. The Wigner-Ville based spectrum sensing [55] derives a greyscale image
of the time-frequency description of the received signal through the Wigner-Ville
transform. Similar to the wavelet based detection, with the aid of edge detection.
2.8. Summary 29
Approach Air-interfaceInformation Synchronisation Computational
Complexity Latency
Filter Bank X X Low Low
Joint M ulti-Band Detection / / High High
Sequential M ulti-Band Detection / X High Low
Wavelet Based Detection X X Medium Low
W igner-Ville Based Detection X / High High
Table 2.2: Summary of the state-of-the-art local sub-band spectrum sensing approaches.
it is able to detect occupied frequency bands. Recently, a multi-band joint detec
tion for spectrum sensing has been introduced in [50], where spectrum sensing is
performed through a class of optimisation problem with the objective of improv
ing the aggregate opportunistic throughput of the opportunistic spectrum access
user while limiting the interference to the other users in the system. However, this
technique requires the knowledge of noise power and the squared values of the chan
nel frequency responses, which makes this approach only practical in fixed wireless
networks, i.e., TV broadcast bands. While [51] investigates multi-band spectrum
sensing algorithm, which supports quality-of-service traffic. In particular, [51] pro
poses a sequential sensing, where a bank of sequential probability ratio tests are
run in parallel to detect the availability of sub-bands, while ensuring a fixed min
imum rate for the opportunistic user. This approach is based on the assumption
that the propagation channel between the transmitter and the spectrum sensing
device is fixed and deterministically known to the opportunistic users, making this
multi-band spectrum sensing approach also only suitable for fixed networks.
2.8 Summary
Spectrum sensing device needs to continuously monitor the spectrum for possible
presence of the vacant frequency bands. In this chapter, we have discussed various
local spectrum sensing techniques that exploit null, minimal, or full knowledge of
the transmitted signal characteristics. We have also addressed the state-of-the-
art local sub-band spectrum sensing approaches, which can effectively improve the
overall spectrum utilisation by simultaneously search over multiple frequency sub
bands at a time. Furthermore, it is realised that existing schemes can hardly meet
the requirements of a fast and accurate spectrum sensing, particularly, in low SNR
range, (considering that the target SNR for a reliable spectrum sensing sensitivity
is about -20 dB [5]) without introducing high complexity to the system. This
2.8. Summary ' 30
observation motivates us to develop a new local spectrum sensing scheme, which
can significantly improve the state-of- the-arts and provides a practical solution.
Chapter 3Cluster-Based Differential Energy
Detection for Spectrum Sensing in
Multi-Carrier Systems
3.1 Introduction
In this chapter a novel spectrum sensing scheme namely, cluster-based differential
energy detection is presented. The proposed spectrum technique has several dis
tinctive features, including low latency, high accuracy, reasonable computational
complexity, as well as robustness to low SNR. The proposed scheme is specially de
signed for sensing multi-carrier sources since most of the current and future mobile
networks are multi-carrier based systems. Hence, it has a wide range of practical
applications.
The key idea of the proposed scheme is to exploit the channel frequency diversity
inherent in high data-rate communications using the clustered differential ordered
energy spectral density (ESD). Specifically, after the ESD computation, the cluster
ing operation is utilised to group uncorrelated subcarriers based on the coherence
bandwidth to enjoy a good frequency diversity. The knowledge of coherence band
width does not need to be very accurate (here we employ the reciprocal of the
maximal channel delay). Furthermore, making use of order statistics of the esti
mated ESD, we further increase the reliability of the sensing algorithm.
31
3.2. System Model and Problem Formulation 32
In order to exploit the second order moment diversity of the observed signal, a
differential operation is performed on the rank ordered ESD. When the channel
is frequency selective and the noise is white, the differential process can effectively
remove the noise floor resulting in elimination of the noise uncertainty impact which
is the main factor making energy detection reluctant [9]. At the final stage of the
proposed scheme, the differential rank ordered ESD within different clusters are
linearly combined in order to further reduce the effect of impulse/spike noise. Binary
hypothesis testing is then applied on either the maximum or the extremal quotient
(maximum-to-minimum ratio) depending on the wireless channel characteristics of
the sensed environment. More importantly, the proposed spectrum sensing scheme
is designed to allow robustness in terms of both time and frequency offset.
In order to analytically evaluate the proposed scheme, both PD and PFA are derived
for Rayleigh fading channels. The closed-form expression shows a clear relationship
between the sensing performance and the cluster size, i.e., channel coherence band
width, which is an indicator of the diversity gain. Computer simulations are carried
out in order to evaluate the effectiveness of the proposed approach and to compare
the performance of the proposed scheme with state-of-the-art spectrum sensing
schemes where up to 10 dB gain in performance can be observed.
3.2 System M odel and Problem Formulation
3.2.1 Multi-Carrier Systems
Transmitted Signal
A general framework of multi-carrier systems has been presented in [57]. The trans
mitted signal can be expressed in the matrix form, Xfc = ^Sk , where is an J x 1
transmitted signal block, Sfc is an M x 1 information-bearing symbol block with the
covariance o-^Im (M stands for the number of subcarriers, and Im for the identity
matrix of size M ), Ÿ is an J x M (J > M) tall pre-coding matrix with full column
rank, and subscript k is the block index. There are two conventional approaches
for implementing the pre-coding matrix Ÿ, i.e.,
C P : ^ (3.1)
ZP: ^ A 0^"^$ (3.2)
3.2. System Model and Problem Formulation 33
where !F is the M x M normilized discrete Fourier transform (DFT) matrix [24],
in (3.2) is formed by collecting the last ( J — M) columns of JF, $ is an M x M full
rank matrix. This work is focused on the cyclic prefix (CP) based system since it has
been widely deployed in practical networks due to its advantages, e.g., eliminating
inter-symbol interference and handling multi-path channels [58]. Nevertheless, it
is shown in Section 4.6.3 how the proposed spectrum sensing scheme can be easily
extended to the zero-padding (ZP) based system in (3.2).
Signal Analysis at the Sensing Device
Consider a wireless device sensing a particular frequency band, in the absence of
the multi-carrier signal, the device can only receive noise, otherwise, it receives a
signal distorted by the frequency-selective channel (denoted by h), timing offset
(denoted by e), frequency offset normalized by the subcarrier spacing (denoted by
e), and additive white Caussian noise (denoted by v). Indeed, there are many other
distortions such as phase noise and non-linear distortions due to imperfect electro
components [58]. However, we will focus on those major physical distortions, (i.e.,
frequency selective channel, noise, timing and frequency offsets) in order not to
diverge the presentation of the key concept.
Given that the spectrum sensing device knows some key parameters of the operating
air-interface such as the block length J , the number of subcarriers M, and the block
duration Tb, the received continuous-time signal is sampled at the sampling period
of Ts = (Tb)/(J). Hence, the timing offset can be expressed into two parts: the
integer timing offset = [(e/7^)J and the fractional timing offset (e — rie), where
[•J denotes the floor operator. It is understood that the fractional timing offset can
be incorporated into the channel impact. Hence, the discrete-time equivalent form
of the received signal is [59]
cYn — ^ ^ y ^t^kJ-\-n—t—Til "bî/n, (3.3)
1=0
where C denotes the upper bound of channel order {C < J — M), O" = exp ,
and the block index k = [(n)/(J)J. Consider an J x 1 vector
Yk = [ykJ+i,YkJ+2,- ,YkJ+j]'^, where (-)^ stands for the matrix transpose, then
3.2. System Model and Problem Formulation 34
(3.3) can be expressed as the matrix form
Yk = flk(A(ne)xk + V(ne)xfc-i) + v&, (3.4)'------V------ '
IBI
where flk — diag{SH(*’ +^), Vk is the corresponding noise vec
tor, A(ne) is a lower triangle channel matrix, and V(ne) is a upper triangle channel
matrix. The detailed layout of both channel matrices depends on the timing offset
He, and the term 'V(ri£)xk-i is the inter-block interference (IBI).
3.2.2 Effect of Second Order Moment
The second-order moments of yk in (3.4) can be computed as below
E (y ty f )=
+ < 7 jV (% )* * " V " (n ,)+ V o Ij , (3.5)
where Afo is the noise power. It is observed that the above result is constant with
respect to the block index k, and the carrier frequency offset (CFO) impact has been
completely removed. This means that the second-order moments of y„ has a period
of J. Furthermore, highly likely the diagonal entries of E(yfcyj^), for n = 1,2,..., J ,
are not constant with respect to the index n due to the frequency selectivity nature
of the communication channels in high data rates.
Remark: In practice, the processing (3.5), i.e., ensemble average, is replaced by the
time average
H y k V k ) (y&yf)' (3-6)fc=0
where K is the number of observation windows and C”' is an J x J matrix. The
above substitution is due to limited processing time available. This would result
in fluctuation of the ESD of AWCN (generation of impulse or heavily tailed noise).
Hence, affecting the performance of any spectrum sensing algorithms regardless of
what scheme is being employed.
3.2.3 Statement of The Spectrum Sensing Problem
The general problem of local spectrum sensing is modelled as the binary hypothesis
testing with hypothesises: Tlo, when the signal is absent; and Hi, when the signal
3.3. Cluster-Based Differential Energy Detection 35
is present.
The specific problem of interest in this chapter is: given the noise to be white
Gaussian, and independent of the multi-carrier signal which is second-order
cyclostationary with the period of N, what is the efficient way to determine the
presence of the signal formulated in (3.3) specifically, in low SNR range?
It is understood that the random sequence x^ due to it cyclostaionairty character
istics has the property E(xfcX^) = E{xk+N^k+N) [24], and the random sequence
Vk satisfies E(vjkVj^) = Afo. In other words, in the absence of signal, i.e.. Ho, the
random sequence of observation {y^} is a white process, otherwise, i.e.. Hi, a second-
order cyclostationary process. Furthermore, as mentioned in Section 3.2.2, making
use of second-order moment of {y^} results in overcoming the CFO phenomenon.
Hence employing second order moment yields the hypothesises
and
where and (a) denote a central Chi squared distribution with 2K degrees
of freedom and a non-central Chi squared with non-centrality factor a, respectively.
Thus, if the SNR of the received signal, i.e., was fairly high, the hypothesis test
in (3.7) will be trivial. The problem of interest in this chapter is to consider spectrum
sensing in very low SNR which, given (3.7), is a rather challenging problem.
3.3 Cluster-Based Differential Energy D etection
3.3.1 Sensing of CP-Based Multi-Carrier Signals
Form an M X M matrix C”' by collecting the last M columns and rows of C”'
defined in (3.6). Consider the special case where the timing offset ric = 0, i.e., Cy.
Due to the effect of CP, the second term at the right hand of (3.5) vanishes, i.e.,
IBI is removed, and the residual term can be written as
cl = + JVoIm, (3.9)
3.3. Cluster-Based Differential Energy Detection 36
Receive^Signal
Order
Order
Order Differentiation
Differentiation
Differentiation
Figure 3.1: Block diagram of the cluster-based differential energy detection algorithm.
where Cm is an M x M circulant channel matrix defined in [57]. Then, an M-point
DFT operation is performed on Cy leading to
c j â :FCy = + JVoI m , (3.10)
where T>m — ^ C m ^ ^ is an M x M diagonal matrix, whose diagonal entries are
in fact the channel frequency response (denoted by hm)- Let 0 ^ be the mth row
vector of 0 . Hence, the mth diagonal entry of Cy reads as
(3.11)
where || • |p denotes the Frobenius norm. Note that, the above expression is equiva
lent to the ESD computation. In many multi-carrier systems, such as OFDM, multi
carrier code division multiple access (MC-CDMA), and single-carrier frequency di
vision multiple access (SC-FDMA), the term ||0 ^ |p is normalized [58]. Therefore,
the ||0 ^ |P term in (3.11) can be ignored leading to
[Cyj = + Vo, m = 1, 2, • • • , M. (3.12)
Based on (3.12), we propose a cluster-based differential energy detection technique
with the following steps. An overview of the proposed technique is illustrated in
Figure 4.1.
81) Group [Cy]m, for m = 1,2, ...,M , into B, where B — M /L , clusters with
3.3. Cluster-Based Differential Energy Detection 37
each cluster having L elements. The mathematical form of each cluster can
be expressed by
- [py]%, Py]t+B ,- , [Cy]i+(L-1)BV', « = 1,2,..., B. (3.13)
The grouping criteria are: cl) elements within each cluster are statistically un
correlated or weakly correlated; c2) all clusters are almost identical or strongly
correlated in the noiseless case, i.e., q i = q2 = ... = qg. The criterion cl) is
to assure that the channel gain within each cluster is sufficiently selective since
the proposed differential energy detection technique aims to take advantage of
the spectrum fluctuation induced by channel frequency selectivity. The crite
rion c2) is mainly for the purpose of de-noising through linear combination of
all clusters on the step S3). Here, the noise is mainly referred to the residual
noise after the second-order statistics (3.6).
In order to fulfil the criteria cl) and c2), we first divide the whole frequency
band into L sub-bands with each accommodating B subcarriers. The mathe
matical form of the Ith sub-band is expressible as:
Pi - [PJ](i-i)B+i>Py](i-i)B+2v I = 1,2,... ,L. When the band
width of each sub-band is smaller than the channel coherence bandwidth, all
elements in p; are highly correlated or approximately identical. Moreover,
we can configure the parameter B such that the bandwidth of the group
[p^, [êy];g+i]^ is larger than the coherence bandwidth such that any two
adjacent sub-bands are weakly correlated or even statistically independent.
With the above configuration to be satisfied, the cluster q can be generated
through block wise interleaving of pz, / = 1,2,..., L.
The above statement implicitly indicates that the clustering process requires
the knowledge of the coherence bandwidth which can be computed assuming
the availability of accurate channel models. In case the accurate channel
models are not available at the sensing device, we can use the upper bound
of channel order C to approximately estimate the coherence bandwidth (for
instance we can let B = [M /£J since the coherence bandwidth is generally
inversely proportional to the channel order). Although, there is no optimal
approach proposed to configure the parameter B, our simulation results in
Section 3.5 demonstrate excellent performance when using the configuration
B = [ M / jC\.
It might also be worth mentioning that the idea of subcarrier clustering has re-
3.3. Cluster-Based Differential Energy Detection 38
cently received a lot of interests particularly for improving the communication
quality and spectral efficiency in cognitive communications [60]- [63]. How
ever, in our work, the subcarrier clustering is for improving the performance
of spectrum sensing.
S2) Sort Qi in an ascending manner, and apply differentiation on each cluster
respectively. This can be viewed as a rank conditioned rank selection pro
cess [64], where the order can change in an adaptive manner from zero to L.
Advantages of such filtering process would be the insensitivity towards heavy
tailed noise and impulsive noise while preserving the edge information [64]-
[66]. The sorting operation allows smoothing of the input without affecting
the statistics of the overall input. Furthermore, the differential operation
allows us to observe the available second order moment diversity.
As it can be observed from (3.12), the sorting function will not have an effect
in 7^0 scenario given that qij'Ho = A/q. When considering a more practical
scenario, i.e., limited number of samples, (ensemble average E(.) replaced by
the time average (3.6)) we will experience noise power fluctuations. Thus,
qil'Ho will no longer be constant and will follow the distribution described
in (3.7). Given that the input signal at this stage, q^, is independent (due
to the clustering operation performed in the previous stage) and identically
distributed, with cumulative density function Fg (q), the probability density
function of the output of the sorting operation is given by [67]
/,.,.{q) = ’- ( ^ ) j ’r * ( q ) ( i - - F i ( q ) ) ' '‘ ’'/ ,(q ) , (3.i4)
where r (1 < r < L) is the rth value returned after the sorting operation, and
/g(q) is the input probability density function. It can be observed from (3.14)
that fq^.^ (q) is the product of the density function of the input, i.e., /g(q),
and the function
»nL(q) = r ( 4 f - ' ( q ) (1 - • (3.15)
It can be concluded that (3.15) is equivalent to beta probability density func
tion [24]. Hence, the sorting operation is equivalent to multiplication of the
input distribution function with a beta function, with shape parameters equal
to r and L — r + l. Replacing u = Fg{q), the expression of the expected value
3.3. Cluster-Based Differential Energy Detection 39
of the value of the output can be calculated using
< r \ r o o
lE(gr:Z<)—
Lr j J-oo
1
q F : - X q ) ( i - ^ g ( q ) r V , ( q )
(3.16)
Wr:I,(u)
where F~^{u) = q (since Fq is increasing in addition to being continuous) and
Wr:L{u) is the sorting function corresponding to rth highest value from set
containing L elements. The above equation reveals that the expected value
after sorting operation is the integral of the product between the sort function,
Wrxiiu), and the inverse distribution function. Figure 3.2 shows the sorting
function and the input distribution superimposed and further demonstrates
how sorting operations allows focusing on a particular region. Thus, the
sorting operation will reduce the effect of noise power fluctuation through
smoothing the sudden changes by focusing on a specific region of the input
density function out one time, this can be particularly useful when dealing
with impulse/spike noise hence, having a direct effect on the error probability.
1.4S o rt Function for r = 7 G um be! Distribution
Rayieigh Distribution Exponentiai Distribution
1.2
0.8
0.6
0 .4
0.2
Figure 3.2: Effect of the sort function on the output, for AT = 50 and r = 7 on various distributions. This figure illustrates how the sort function focuses on a particular point of a distribution based on the value of r. Furthermore the shape difference for various distribution all having a mean value of 0.42 is also shown in this figure.
The sorting problem has attracted a great deal of research and since early
3.3. Cluster-Based Differential Energy Detection 40
1950s many sorting algorithm have been introduced in the literature, e.g.,
bucket sort, counting sort, spread sort. A comprehensive description of various
search algorithms can be found in [68]. Hence, sorting operation in this step
can be implemented using one of many developed sorting algorithm based on
the memory/efficiency trade-off the spectrum sensing device requires. There
fore, the device does not need to perform the operations explained in (3.14)-
(3.16) to sort the data.
The main objective of the differential operation, which is further performed
in this stage, is to remove the constant noise floor, i.e., A/q, contained in all
elements. The output of differentiation is denoted as with its Zth element
given by
[q<]i = I (3.17)
It is clear that [q ][ is zero for all I in the absence of the signal, and under
goes a fluctuation in the presence of the signal due to the channel frequency
selectivity. This distinctive feature motivates the test statistics presented in
S3) and allows us to overcome the noise uncertainty problem inherent in the
conventional energy detection.
Furthermore, this stage is intended to exploit the second order moment diver
sity of the input signal distribution. Figure 3.2 illustrates the shape/feature
difference [69] (in terms of inverse CDF) which exists between various dis
tributions. All three distributions in this flgure have equal mean value, yet
regions exist where the distributions are very distinct from one to an other.
In the case of no shape/feature difference, the performance of the proposed
technique will degrade. Since today’s high data rate communications always
leads to frequency selective channel, we will experience shape difference and
consequently second order moment diversity.
S3) Perform linear combination of q for i = 1 , 2 , . . . , # for the purpose of de-
noising, and then the following test
Test I : max
maxTest II : -----
B
' (3 18)I
mm
t=i
i E L o ' i1 A2, (3.19)
3.3. Cluster-Based Differential Energy Detection 41
where the threshold Ai, A2 should be carefully configured to manage the PD
and PFA, which will be discussed in the performance analysis (see Section IV).
The test metrics presented in (3.18) and (3.19) represent the maximum and
the maximum to minimum ratio of the clustered ESD respectively, which have
been widely used for sub-optimum decision with low computational cost [70].
It is shown in Section IV-C that the proposed differential energy detection tech
nique can offer comparable performance to the optimal detector in Neyman-
Pearson sense [71], however, the latter requires the knowledge of channel gain,
noise power and signal power, which are often not available in practice for the
spectrum sensing application.
3.3.2 Overcoming Timing Offset
As mentioned in Section 3.2.2 the effect of CFO has been already solved through
employment of second-order statistics. Now, our main concern is to overcome the
timing offset. In fact, the special case of rie = 0 can be hardly captured due to the
lack of timing synchronization mechanism before the spectrum sensing component.
In order to handle the problem of unknown timing offset effectively, we propose
an “one ballot veto" policy to reject the hypothesis H q. The policy is stated as
follows:
51) Form J x 1 vectors, = [ykJ+i+ô,ykJ+2+ô,- ,ykJ+J+ô]'^, k=o,i,...,K, where
Ô denotes the offset in time,
52) Compute A E{yk,57^,5) according to (3.5), for
(J = 0 , (J - M), 2 (J - M ) , . . . , M;
53) Apply the cluster-based differential energy detection explained in Section 4.4.3
on V . If for any value of 6 the test statistic satisfies Hi criterion
it is understood that the signal is present and the cluster-based differential
energy detection algorithm would not be applied on the input after detecting
the first value of meeting the Ho condition.
The underlying idea is, in the presence of a signal, there exists such a S fulfilling the
condition | ne— < J —M, and under this condition, the proposed spectrum sensing
scheme can successfully reject the IBI. In the absence of signal, is approx
imately constant with respect to 5, due to constant energy of AWGN throughout
3.3. Cluster-Based Differential Energy Detection 42
the spectrum. Most certainly, this stage will add to the overall complexity of the
algorithm which would be shown in Section 3.4.4. However, in order to increase the
reliability of the sensing device, implementation of this stage is necessary.
3.3.3 Extension to the ZP-Based System
Let us start from the special case of 71 = 0. Using the result in [57], we can
easily justify that the second term at the right hand of (3.5) vanishes due to the
implementation of ZP, i.e., (3.2). Therefore, (3.5) can be expressed by
E (yityf )= < 72A (n ,)**"A '^(n ,) + M o h (3,20)
= + V olj. (3.21)
Performing J-point DPT on (3.21) yields
:^./E(yfcyf ) : f " = + jVoIj, (3.22)
where $ = J~j is an J x J DPT matrix normalized by the factor (1)/(V J),
C j is an J X J circulant channel matrix with "Dj formed by the corresponding
channel frequency response. It is easy to observe that (3.22) has the same form as
(4.23). Therefore, the three step spectrum sensing algorithm proposed in Section
4.4.3 for the CP-based system can, be straightforwardly, applied on (3.21).
Furthermore, the “one ballot veto” policy can be applied on the ZP-based system
to handle the problem of unknown timing offset.
3.3.4 Knowledge of Key Parameters
The proposed spectrum sensing technique requires the knowledge of several key pa
rameters about the operating air-interface as well as channel models (i.e.,, the block
length J , the number of subcarriers M, the sampling rate Tg, as well as the upper
bound of channel order C). Those knowledge of parameters are very commonly
assumed in almost all estimation and detection techniques including spectrum sens
ing, e.g., in [11] [25] [50] [49]. Lack of the knowledge of these parameters would
result in performance degradation for all spectrum sensing techniques. Practically,
it is possible to obtain the mentioned parameters through accessing a geo-location
database. For example, the new Ofcom regulations [7] allow for sensing devices to ac
cess location-aided databases for obtaining key parameters about local air-interfaces
3.4. Performance Analysis 43
and channel power delay profiles (PDPs). Design and maintenance of location-aided
databases is an ongoing research activity in both Europe and US [4], [72]. Surely, the
impact of imperfect knowledge of air-interface parameters on the spectrum sensing
performance is of interest to telecommunication engineers.
3.4 Performance Analysis
Conventionally, the metrics of interests for performance evaluation of spectrum
sensing are mainly the PFA, PD, and computational complexity. The PFA is often
formulated for the AWGN case since it would not be affected by the channel fad
ing. However, the PD is related to the channel fading behaviour, and here we are
interested in the Rayleigh fading scenario. In addition to the PFA and PD analysis,
we will present numerical results as well as the computational complexity of the
proposed approach.
3.4.1 Probability of False Alarm
Let’s consider the special case of Ue = 0. It is understood that elements of q%
(see (3.13)) under the hypothesis Ho follow independent and identical central Chi
squared distributions with 2K degrees of freedom [34], i.e.,
exp (-a /2 ) (3.23)
where F(.) represents the Gamma function [24]. Hence, after the differentiation (ig
noring the effect of the sorting operation), the I th element of qj based on Appendix
A follows the p.d.f.
(3.24)
Remark: In the derivation of (3.24), we ignored the effect of the sorting operation.
This is mainly because the exact probability density function of the r th order statis
tic from any continuous population is rather difficult to deal with (see (3.14)) and
in most cases requires numerical evaluation of a nontrivial integral [67]. Since the
earliest known bounds for the expected value of highest order statistic with was
derived by Gumbel, Hartley and David, much work has been done on statistical
properties of order statistics, the summary of which can be found in [67]. Despite
3.4. Performance Analysis 44
all the work carried out in the area of the order statistics still the only effective way
for determining the distribution of /r:L(q) would be evaluating them numerically.
However, using the probability-integral transformation we are able to approximate
the variance of the r th order statistic, of any continuous distribution as
4 ^ (9 ) - (x + i]2’|j;+2)^ (E^hr.1.]))-", (3.25)
where E[qr-.L]i or in other words the expected value of rth order statistics, can be
approximated by:
(3-26)
where denotes the inverse cumulative distribution of the input signal. Please
note that the above approximations will converge as L ^ oo (see [67, Chapter 3]
for proof). The above approximations indicate that the sorting operation will have
a direct effect on the performance of the proposed algorithm since it will reduce
the variance of the data significantly. Thus, it can be concluded that the sorting
operation will reduce the effect of noise power fluctuation resulting from the limited
observation length. Hence, having a direct effect on the error probability as the test
statistic is subject to less variation. Since it is not mathematically feasible to derive
the performance incorporating the sorting operation we have shown the effect of
the sorting operation in Section V through simulations.
The linear combination q[ = l]^i[q^]z employed in (3.18)-(3.19) will result in the
following moment generating function (MCF) [24]
Ai([q'];|Mo)= n ^ ( |q ; i l l % ) = ((1 - ‘i t y " ) ^ ■ (3.27)
It can be observed that the random variable [q']i|'#o has an Erlang distribution [24]
with the shape and rate parameter equal to w = K B and rj = 0.5. Hence, its p.d.f
is given by
/[q'ldWoH = ex p (-7/0') (3.28)
Accordingly for Test II (see Appendix B), we can derive the p.d.f. of [q']f/[q%|'#o,
V 1 < Z, j < # and j ^ I, bearing in mind that the values of q are non-negative.
3.4. Performance Analysis 45
as [24]
roo rq[z
/[q']j/[q']il«o(^)= / / /[q']i,[q'bl«o dg'-Jo Jo
rOO
= //[q'Ii.Iq'Jil o g!) dgjyW —' r ( 2ro)
Finally, we can obtain the PFA as
(3.29)
Test I : PFA = 1 - (3.30)
Test II :PFA =, A AfT(2ro)2Fi([tî7,2ro],B + l , - A 2) V - V -------------------
where if) = (g), Q{.,.) is the lower Gamma incomplete function, and 2F i([a ,6],c,d)
is the Gauss hypergeometric function [73].
The PFA formulas above indicate the probability where the second order moment
diversity observed from the noise only input is higher than the test statistic. It can
be observed from (4.5) that Test II can only be applied and is meaningful if the
channel order is, L > 3. Hence, given the maximum channel order, one can choose
which test to employ. Furthermore, it can be concluded from (3.30) and (3.31) that
the PFA of proposed schemes is a function of the cluster size L, the number of
clusters, B, and sample complexity K, as well as the thresholds Ai, A2. Specifically,
it is exponentially related to the inverse of the channel delay, i.e., L, implying
that the performance is exponentially eflFected by the frequency selectivity of the
environment. This was expected as the key idea behind the proposed spectrum
sensing approach is to make a decision based on the observed second order moment
diversity resulting from the frequency selective channel. Furthermore, PFA will be
reduced dramatically as AT —>• 00. Given that for practical applications, the PFA is
often given a fixed value, such as 10% as per the FCCs requirement [4], (3.30) and
(4.5) can be employed to determine the appropriate thresholds Ai, A2 for a given air
interface, channel order and the required observation length, i.e., F (A) = 1 — PFA.
The exact effect of threshold value on the performance of the proposed approach is
shown in Section 3.4.3.
3.4. Performance Analysis 46
3.4.2 Probability of Detection
It has been proved that the random variable q,i\Hi follows non-central Chi squared
distribution with the p.d.f. [11]
K -1
W — ^ ^ ^ (3.32)
where X{.) denotes the modified Bessel’s function of the first kind, and j i the SNR
affecting the value.
Furthermore, we consider an interesting case when the SNR, 7 , follows an indepen
dent and identical exponential distribution
/y(o:) = - exp ^ , (3.33)
where 7 denotes the SNR mean.
Remark: In fact, modelling the SNR as an i.i.d. exponential distribution implies
that the communication channel is a Rayleigh fading channel. Rayleigh fading is
considered as one of the most practical models for tropospheric and ionospheric
signal propagation as well as for the effect of heavily built-up urban environments
on radio signals. Rayleigh fading is mostly applicable when there is no dominant
propagation along a line of sight between the transmitter and receiver [8]. Since,
based on FOG regulations [4] there is no guarantee that there would exist a line
of sight between the sensing device and the transmitter, it would be a reasonable
assumption to model the fading channel as Rayleigh fading.
The distribution of A7; = 7; — 7f- i, whose MGF is given by
Af(A7() = — p—. (3.34)1 + W
Hence, it can be concluded that A7f follows a Laplace distribution [24]. Considering
that q'il'Hi follows a non-central Chi square distribution with 2K degrees of freedom
and the non-centrality factor of 2A7/, and also the fact that A7 is non-negative,
the term in Appendix B is computed using the following
PDJ^ = = / Qk (^ , \ A i ) exp dA'yi (3.35)7Ja7!=o ' V 7 /
with (f = 2A7/, where Qk {cl, 6) denotes the generalised Marcum Q-function defined
3.4. Performance Analysis 47
by1 ro o / 2 I -2 \
QK{a,b) = J lK -i{ax)dx . (3.36)
The PD for Test II can be evaluated using Appendix C, where the p.d.f. of
A7i/A 7d^i|?/i given by
pOO
yA 7 ; /A 7 d ^ ; |% i (( ) /J 3 = 0
(l + a )2 -(3.37)
Hence, the term P D ^ in Appendix B given Rayleigh fading is given by
Once more considering the special case of = 0, after the differentiation under
the hypothesis Hi, the differential SNR A ji corresponding to (see (3.13))
follows the Laplace distribution with the p.d.f. based on the derivation in (3.27).
Furthermore, the average of differential SNR A7 ; can be computed by
Then, the term PDJ^ can be evaluated by
X exp dA7 f.
A ji7
xu—l
Based on the analysis in Appendix C, we can further write (3.40) into
PDT = _ _ exp ( _ I X2 + 7 \ 2 + 7
1 + 3 2(24-7)7 V 7
2 4-7Ai 7
2(2 + 7)
+ 2 + 7K -l _
exp ( - y ) X
(3.39)
(3.40)
3.4. Performance Analysis 48
where iF i(.;.;.) denotes the hypergeometric function [73], and Ln{.) the Laguerre
polynomial function defined by
r=l(3.41)
We can obtain the PD for Test I by applying (3.41) into Appendix B.
Evaluating the PD of Test II requires the p.d.f of the ratio of A'y^/A'yjl'Hi. Based
on the derivation in (3.29), we have
-^T{2w)/A7i/A7, |7£i (^) (1 + ^) (3.42)
Then, the term PDf^ can be computed by
PJJT2 _ r ( t o ) (y , V ^ ) dA7 ,. (3.43)
Considering considerably Low SNR such that 1 A7 , the integration in (3.43) can
be computed by using Appendix C and the analysis in [74, Appendix A]. Hence,
P D ^ can be expressed by:
P D P = ®exp ( ^,fc=0
+$exp ( - y ) E iF i (ro; & + l ; y ) (3.44)
where 0 = (the full proof can be obtained by using [75, Eqn. (25)]). Finally,
we can obtain the PD for Test II by applying (3.44) into Appendix C.
It can be observed from (3.41) and (3.44) that the performance of the proposed
spectrum sensing technique, in terms of PD, is affected by the average SNR value 7 ,
sample complexity K and the threshold value Ai and Ag and further exponentially
effected by the channel order L. Moreover, it can be observed that the performance
of Test II improves much faster with the increase in channel order, L. The effect of
various parameters on the PD of the proposed approach will be discussed in detail
and illustrated pictorially in Section 3.4.3.
3.4. Performance Analysis 49
3.4.3 Numerical Results and Discussions
In this section, numerical results based on the PFA and PD expressions found
the Sections 3.4.1-3.4.2, are provided to visually demonstrate the effect of various
factors. Figure 3.9 illustrates how PD is affected by the observation length (latency)
in Test I. The results are generated for the configuration where the number of sub
carriers M = 64, and the number of clusters B = 6. The threshold Ai was fixed
for achieving PFA = 10% with the noise uncertainty factor set to 2 dB (the noise
uncertainty factor in practical scenarios is typically between 1 to 2 dB [9]). The
main factor causing noise uncertainty is the temperature variations at the receiver
which leads to inaccurate noise power measurements. The uncertainty is created
by fixing assumed/ estimated noise power based on the SNR value mentioned, while
the real noise power varies with each realization by a certain degree according to
the uncertainty factor. It is observed that the proposed approach features fast
convergence rate. For example observing the point of PD = 90%, the PD improves
by 5 dB in the SNR when the number of multi-carrier symbols K varies from 3 to
5, while this improvement is as small as approximately 1 dB when K varies from
20 to 30.
Figure 3.10 shows how the channel length C would infiuence the PD when the
observation length is set to K = 10. Take the point PD = 90% as an example, 8 dB
gain in the SNR can be observed when C varies from 0 to 4. Furthermore, 10 dB
improvement when it varies from 4 to 12. It is an interesting result which clearly
indicates the channel frequency-diversity gain inherent in the proposed spectrum
sensing scheme.
The complementary receiver operating characteristic (ROC) curve for both Test
I and Test II (in Rayleigh fading channel) are shown in Figure 3.11 and Figure
3.12 respectively. These Figures refiect a fundamental tradeoff between PFA and
PD. Furthermore, the effect of the threshold value on both PFA and PD can be
also observed, since different threshold values were employed to produce the PFA-
PD tradeoff. In order to have a benchmark and also for performance comparison,
the ROC curve for conventional energy detection with various uncertainty factors
(U) are also illustrated. It is observed that the performance of the energy detection
severely degrades as the uncertainty factor is introduced (this phenomenon has been
fully investigated in [9]). While, due to differential stage of the proposed technique,
it is considerably robust to uncertainty factor. For the sake of comprehensive perfor-
3.4. Performance Analysis 50
mance comparison, Figure 3.11 also illustrates the ROC of the optimal detector in
Neyman-Pearson sense [71]. It should be noted that the optimal detector requires
channel gain, noise power and the transmitted signal power (which is not a feasible
solution in practical scenarios). Hence, as expected it delivers better performance.
3.4.4 Computational Complexity
The main complexity of the proposed scheme is due to the following stages:
1. The second-order time average: for the case of rig = 0, this stage requires
X J complex multiplications and additions.
2. Discrete Fourier Transform: M —point DFT is implemented which introduces
the complexity by 0 (Mlog(M)).
3. Sorting: there are B clusters consisting of L elements, hence, the complexity
of this stage is BLO{L).
4. Differentiation: this stage consists of subtracting every element of from its
previous one for each cluster, hence the computational complexity is given by
BO{L).
5. Linear combination: This would add a further complexity of 0{B).
6. Decision making: Finally the extreme value(s) is selected and compared to
the predetermined threshold value. Consequently adding a complexity factor
of 0{L).
Resulting in the overall computational complexity:
0 { K ‘ J) + 0{M\og{M)) + B{L + 1)0{L) + 0{B) + 0{L) (3.45)
Note that the above complexity is for the case = 0. When employing the “one bal
lot veto” scheme for arbitrary ng (see Section HI-A), the computational complexity
is increased by a factor of ((M )/(J — M)) (in the worst case scenario).
The last three terms are negligible in (4.39), hence, the overall complexity of the
proposed scheme is approximately M [O(A’ J) + 0(M log(M ))] / { J — M). This
reflects that the proposed scheme requires a relatively low computational complexity,
making it suitable for practical scenarios, where computational efficiency is a key
issue.
3.5. Simulation Results and Discussions 51
3.5 Simulation Results and Discussions
Computer simulations were performed to evaluate the proposed spectrum sensing
scheme. The system investigated in this section has M = 2,048 sub-carriers with
the sub-carrier spacing of 15 kHz (3GPP LTE-advanced system [76]), each frame
consists of 7 OFDM blocks with the CP length oi J — M = 160, the sampling
frequency is the same as the signal bandwidth of 30.72 MHz. The carrier frequency
is also set at 5 GHz. The communication channel is generated according to the
WINNER channel model under B2 outdoor scenario [77], and the sensing device
is moving at the speed of 3 km/h. The SNR is defined by the average received
symbol energy to noise ratio at the sensing device. The threshold for hypothesis
test is carefully chosen so that the PFA is fixed to 10%. All simulation results were
obtained by averaging over 2,000 Monte Carlo realizations.
Experiment 1
The objective of this experiment is to examine the analytical analysis obtained in
the previous sections by comparing them against the simulation results based on
the configuration explained above and further to show the effect of the sorting op
eration on the performance of the proposed scheme. Figure 3.7 and Figure 3.8
demonstrate the probability of detection for different observation lengths given var
ious average SNRs for Test I and Test II, respectively. We can observe a very small
difference between analytical results and simulation results when the observation
length is larger than two symbols duration. The difference becomes large when the
observation length is less than two symbols duration. This is mainly caused by the
insufficient statistics used in signal processing. Comparing Figure 3.7 and Figure 3.8
verifies that Test II outperforms Test I, particularly, when the observation length
is short. This difference is mitigated with the increase of observation length. We
have also shown the effect of the sorting operation through simulations in Figure
3.7 and Figure 3.8. As it can be observed, the sorting operation can improve the
performance as the observation length is increased. This was expected as previously
explained in Section 3.3.1 and Section 3.4.1.
3.5. Simulation Results and Discussions 52
Experiment 2
The objective of this experiment is to examine the proposed scheme with respect
to the state-of-the-art spectrum sensing approaches. Since, the proposed approach
is based on exploiting the second order moment frequency diversity, it is essen
tial to check how much gain is introduced due to this exploitation by having the
frequency-domain energy detection as a benchmark for performance comparison.
The threshold setting for energy detection can be found in [11]. The simulation
performed for energy detection are based on noise uncertainty factor, U = 0,1,3
dB and the threshold is based on the assumed/estimated noise power while the real
noise power varies with each Monte Carlo realization by a certain degree depending
on the uncertainty factor. Figure 3.13 shows the performance comparison when ob
servation length, K = 7 symbols. It can be observed that the performance of energy
detection is considerably dependent on the noise uncertainty factor. It is further
proved in [9] that increasing the observation length does not affect the performance
of the energy detection scheme when the the exact noise power is not known, i.e.,
U 0. Figure 3.13 also illustrates the performance of the second order cyclosta-
tionarity based detection. The proposed approach is able to outperform the second
order cyclostationarity by at least 8 dB when K = 7. Cyclostationarity based de
tection relies on the cyclic frequency of the received signal to determine existence of
a source. Hence, deep fading at cyclic frequency can have a detrimental effect on its
performance while the proposed technique takes advantage of this fading to exploit
the moment diversity. The performance of the proposed approach is also compared
to Wigner-Ville based spectrum sensing [55]. For this purpose, in order to have fair
comparison, we have modified the original work in [55] to accommodate a SISO en
vironment. As shown in Figure 3.13 the mentioned approach can deliver acceptable
performance up to SNR of —16 dB. However, as the SNR further decreases, the
performance of Wigner-Ville transform based approach also decreases. This was ex
pected since noise power fluctuation increases such that it makes the edge detection
used in this approach reluctant. The performance comparison is also carried out
for pilot based detection [78] and differential based energy detection [79]. In order
to carry out simulations for the pilot based detection, it is assumed that the pilot
symbols are embedded in each OFDM block, which are equally spaced for every
16 or 32 sub-carriers. Since, the mentioned pilot based detection is based on the
energy of the pilot symbols it is affected by the noise uncertainty factor. Therefore,
the performance was evaluated for Z7 = 0 dB and [7 = 2 dB. Figure 3.13 shows that
3.5. Simulation Results and Discussions 53
the proposed technique is able to outperform the mentioned technique by at least
10 dB. It should be also noted that pilot based detection requires synchronization
and pilot information while this problem can be overcome in the proposed technique
with the implementation of the “one ballot veto” policy.
Since the detection technique introduced in [79], also exploits the frequency diver
sity of the channel our main concern is to observe how much improvement can be
delivered by the clustered-based energy detection. Figure 3.13 further shows that
the proposed approach can outperform the differential energy detection by at least
5 dB in low SNR environments. This improvement is mainly due to the clustering,
the linear combination and the “one ballot veto” policy which are implemented. It
is noteworthy to mention that the clustering operation not only improves the per
formance of the proposed techniques but also reduces the complexity by a factor of
0(2M ).
Experiment 3
The objective of this experiment is to observe the performance improvement due
to the clustering and differential stages. As it was explained, the main purpose of
the differential stage is to remove the AWGN which is available in all the frequency
bands and to further exploit the frequency diversity, while the clustering operation
is to remove any possible correlation in the ESD due to the fading channel. In
order to observe how much improvement can be achieved when incorporating these
two stages (i.e., clustering and differential stages), we have set an experiment were
the proposed technique in Section 4.4.3 is compared to its equivalent without the
mentioned two stages. Without differential and clustering operations the proposed
technique can be thought of as a simplified eigenvalue detection [35] where instead of
making decision based on the ratio of the eigenvalues of the covariance matrix of the
received signal, the decision is based on the ratio of the maximum and the minimum
of the ESD of the received signal. This comparison is possible since, in multi-carrier
systems, parallel transmission is performed, hence, the DFT decomposition can
be considered as a special case of eigenvalue decomposition. The result of this
performance comparison for different observation lengths is shown in Figure 3.14.
As it can be observed we are able to achieve up to 10 dB gain in performance, this
gain is more apparent as the observation is increased.
3.6. Summary 54
Approach Air-interfaceInformation Synchronisation Computational
Complexity Latency
Energy Detection X X Low Low
Cyclostationarity Detection / / High High
Covariance Based Detection X X Medium Low
Eigenvalue Based Detection X X High Medium
Matched Filtering / / Low Low
Pilot Based Matched Filtering / / Low Low
Entropy Based Detection / X Medium Medium
Kurtosis Based Detection / X High High
Clustered-Based Energy Detection / X Medium Low
Table 3.1: Comparison of the state-of-the-art local narrow-band spectrum sensing approaches and the cluster-based energy detection.
3.6 Summary
In this chapter, a novel differential energy detection scheme for multi-carrier sys
tems, namely, cluster-based differential energy detection, which can form fast and
reliable decision of spectrum availability even in very low SNR environment, has
been proposed. It has several distinctive features including low latency, high accu
racy reasonable computational complexity, as well as robustness to very low SNR.
For example, the proposed scheme can reach 90% in probability of detection and
10% in probability of false alarm for an SNR as low as —21 dB, while the observation
window is equivalent to 2 multi-carrier symbol duration. The proposed scheme at
this stage is specially designed for sensing multi-carrier sources but we would argue
that since most of the current and future mobile networks are multi-carrier based
systems, that the proposed scheme has wider ranging practical applications. The
proposed approach can deliver desirable performance in high density communication
networks and urban environments due to its robustness in low SNR environments.
Furthermore, the clustered-based differential energy detection can be employed in
vehicular communication systems.
The key idea of the proposed scheme is to exploit the channel frequency diversity
inherent in high data-rate communications using the clustered differential ordered
energy spectral density. Initially, the ESD of the received signal is estimated. Follow
ing the ESD computation, the clustering operation is utilized to group uncorrelated
subcarriers based on the coherence bandwidth to enjoy a good frequency diversity.
The knowledge of coherence bandwidth does not need to be very accurate (we em
ploy the reciprocal of the maximal channel delay). Furthermore, making use of
order statistics of the estimated ESD, we increase the reliability of the sensing al
3.6. Summary 55
gorithm. This will allow us to smooth the fluctuation in noise ESD, resulting from
limited observation length which affects the statistics of the received signal. Hence,
this will stage will have a direct effect on the PFA of the proposed approach.
In order to exploit the second order moment diversity of the observed signal, a
differential operation is performed on the rank ordered ESD. When the channel is
frequency selective and the noise is white, the differential process can effectively re
move the noise floor resulting in elimination of the noise uncertainty impact which
is the main factor making energy detection reluctant [9]. Furthermore, the differ
ential stage will allow us to exploit the frequency selectivity which available in the
received signal. At the flnal stage of the proposed scheme, the differential rank
ordered ESD within different clusters are linearly combined in order to further re
duce the effect of impulse/spike noise. Binary hypothesis testing is then applied
on either the maximum or the extremal quotient (maximum-to-minimum ratio) de
pending on the wireless channel characteristics of the sensed environment. More
importantly, the proposed spectrum sensing scheme is designed to allow robustness
in terms of both, time and frequency offset, without compromising computational
complexity. Additionally, it is worth mentioning that given not a very frequency
selective environment, i.e., when the transmitted signal is experiencing flat fading,
the performance of the proposed scheme is degraded to the performance of the
frequency energy detection with approximately zero uncertainty factor.
To analytically evaluate the proposed scheme, both PD and PFA were derived for
Rayleigh fading channels. The closed-form expression showed a clear relationship
between the sensing performance and the cluster size, i.e., channel coherence band
width, which is an indicator of the diversity gain. Computer simulations are carried
out in order to evaluate the effectiveness of the proposed approach and to compare
the performance of the proposed scheme with state-of-the-art spectrum sensing
schemes where up to 10 dB gain in performance can be observed. This would imply
that employing the proposed approach in a communication system will make the
network of interest more robust to hidden node problem, i.e., mitigate interference
to heavily shadowed licensed users. More interestingly, it has been shown through
simulation that the proposed spectrum sensing algorithm has a long convergence
time, which allows it a suitable in delay sensitive systems.
3.6. Summary 56
0 .95
0 .9
0 .85
IS 0.8
0 .75K = 3 K = 5•9
S 0 .72Q.
0 .65 K = 20 K = 30
0 .5
SN R (dB)
Figure 3.3: The relationship between the PD and the observation length for M = 64 and £ = 6 .
A -0.9
0.8
S 0 .7
0.5
Q- 0 .4 — Fl at F ading - A - C h an n e l Length= 2
^ C han n e l Length = 4 ' - D - ' C h an n e l Length = 8 — O — C h an n e l Length = 12
0 .3
S N R (dB)
Figure 3.4: The relationship between the PD and the coherence bandwidth, C, and the observation length K =\6.
3.6. Summary 57
- A - P ro p o se d T e chn ique Ti K = 6 ' P ro p o se d T e chn ique TI K = 10
' " O ' " P ro p o se d T e chn ique TI K = 15 • -O" ' P ro p o se d T ech n iq u e TI K = 20 E nergy D etection U=2 dB , K= 10' - ' - ' E nergy D etection U=1 dB, K= 10 ' E nergy D etection U=0 dB, K= 10- - - O ptim al D etector, K =6 — — — O ptim al D etecttor, K=10
0.01 0 .02 0 .03 0 .04 0 .05 0 .06 0 .0 7 0 .08 0 .09 0.1Probability of F a ise Alarm
Figure 3.5: Complementary ROC curves of the Test I and it’s comparison with energy detection for various uncertainty factors (U), the optimal detector based on Neyman-Pearson criteria. 7 = —lOdB, C = 8 and M = 64 based on the analytical results in Section 3.4.
Q 0.8
= 0 .75 -A - P ro p o se d T ech n iq u e Til K P ro p o se d T ech n iq u e Til K P ro p o se d T ech n iq u e Tii K P ro p o se d T echn ique Til K E nergy D etection U=0 dB, E nergy D etection U=1 dB,
E nergy D etection U=2 dB,
0 .02 0 .0 3 0 .04 0 .05 0 .06 0 .07Probabiiity of F a ise Alarm
Figure 3.6: Complementary ROC curves of the Test II and it’s comparison with energy detection for various uncertainty factors (U). 7 = —lOdB, £ = 5 and M = 64 based on the analytical results in Section 3.4.
3.6. Summary 58
o 0 .7 -A - Cyclostionarity D etection P ro p o se d T ech n iq u e T e s t I P ro p o se d T echn ique T e s t li
■ - O * ■ E nergy D etection U = 3 dB■ - 0 - ' E nergy D etection U = 0 dB • ' E nergy D etection U = 1 dB- O - Pilot B a se d D etection U = 2 dB
Differentiai E nergy D etection W igner-V iie b a s e d D etection
- 2 5 - 2 3 -2 1 - 1 9 - 1 7 - 1 5 - 1 3 -1 1 - 9S N R (dB)
Figure 3.7: The performance comparison of the proposed technique, frequency-domain energy detection, second order cyclostationarity, pilot based detection and differential energy detection for K = 7.
A , À - - é T ^ '
- n - K= 2 OFDM S ym bois- O - K= 7 OFDM S ym bo is
A - K= 14 OFDM S ym boisK = 2 OFDM S ym boi (Differntiai)K = 7 OFDM S ym bo is (Differentiai) K = 14 OFDM S ym bo is (Differntiai)
-2 5 - 2 3 -2 1 - 1 9 - 1 7 - 1 5 - 1 3 -1 1 - 9 - 7 - 5SN R (dB)
-3 -1 1
Figure 3.8: The effect of the differential and clustering stages on the performance of the proposed spectrum sensing technique.
Chapter 4
A Bayesian Model Based Approach
for Joint Sub-Band Level Spectrum
Sensing
4.1 Introduction
The problem of model estimation/ selection for array processing has been well inves
tigated in the literature. For example, the approach developed by Lawley [80] and
Bartlett [81] using a sequence of hypothesis tests with subjective thresholds for each
test, the model selection developed by Akaike [82], Schwartz [83], which addresses
this problem by selecting the model which results in minimum information criteria
using the log-likelihood of the maximum likelihood estimator of the parameters in
the model. Model estimators based on the eigenvalue of the covariance matrix was
further proposed by Kailath and Wax [84], namely, the Kailath-Wax Akaike infor
mation criteria estimator and Kailath-Wax minimum descriptive length criterion
estimator. Unfortunately, due to high computational complexity and sample size
requirements of the available approaches, they are not able to fulfil the demand
ing requirements of practical spectrum sensing for opportunistic spectrum access
applications, e.g., cognitive radio.
The main contributions of this chapter are 1) re-defining the objective of the sub
band level spectrum sensing device to a model estimator, 2) deriving the optimal
model selection estimator for sub-band level spectrum sensing for fixed and variable
59
4.2. Problem Formulation 60
number of users along with a sub-optimal solution based on Bayesian statistical mod
elling and 3) proposing a practical model selection estimator with relaxed sample
size constraint and limited system knowledge for sub-band spectrum sensing appli
cations in OFDM A systems. The proposed technique takes advantage of the second
order moment channel frequency diversity. More interestingly, it does not require a
priori knowledge of noise power and the propagation channel gain, and is designed
in such a way to show robustness towards energy leakage. The proposed model
selection based sub-band level spectrum sensing approach is analytically evaluated
through probability of false alarm and probability of detection along with closed
form expression for the threshold value. Furthermore, computer simulations are
carried out in order to evaluate the effectiveness of the proposed scheme.
4.2 Problem Formulation
Consider a communication system operating over a wide-band channel that is di
vided into K non-overlapping sub-bands, e.g., multi-carrier systems. However, in
a particular geographical region within a certain time frame only I number of the
sub-bands is utilized by the users, where I < K . Thus, [K — I) sub-band are avail
able for opportunistic access. The essential task of the spectrum sensing device is
to determine the availability of these {K — Ï) sub-bands.
The estimated power spectral density (PSD) [24] of the received signal, sampled at
frequencies {fk}k=v cem be modelled as
(fk — oikSk + Vk + jk , l < k < K (4.1)
where Sk denotes the power of the signal transmitted at the sub-band, ak is
an indicator function taking a value of 0 if the k^^ sub-band is vacant or 1 if it
is occupied by a user, Vk is the additive white Gaussian noise (AWGN) power at
the k^^ sub-band, and 7fc denotes the PSD estimation error due to limited obser
vation length at the k^^ sub-band. Please note that the effects of various physical
impairments in a communication systems is further discussed in Section 4.4.2.
Hence, (4.1) leads us to consider inferring the vacant sub-bands in the bandwidth
of interest, converts the spectrum sensing problem to a model estimation problem
where the parameter of interest is a = [o;i,a;2, . . . ,o:/ï-]^. Based on the model
above the task of an ideal spectrum sensing device, given ot is the estimation of a .
4.3. Optimal and Sub-Optimal Solutions 61
is to select the binary sequence ex. f o r which the a posteriori probability
distribution Pr{a.\(p), where (p = - - -, is m a x im u m over all
the possible binary sequence o f o l .
4.3 Optimal and Sub-Optimal Solutions
4.3.1 Optimal Decision Rule
Based on the problem stated in Section 4.2, there are at most 2^ sequences which
can be associated to a . Assume that the transmitted signal in each sub-band has
a unit power, i.e., = 1. This assumption holds when the users adopt a
uniform transmission strategies given no channel knowledge at the transmitter side.
Considering that each sequence is equiprobable using Bayes rule [85] we have
Based on the above equation, in order to maximize P(â|<^), our objective is to
maximize P {ip\ôc), given that
P(V5|a) = y 'p ( y ) |â ,7 ) P ( 7 )d 7 . (4.3)
Moreover, based on the variable definitions in Section 4.2, we have
/ 1 \ ^ / 2 f I JL \P ( v |â ,7 ) = j exp ( - * - 7i) j , (4.4)
where denotes the noise variance. For the sake of reaching a conclusion it is now
necessary to specify the distribution which 7 follows, hence, we will assume jk is a
random variable following a Gaussian distribution with variance of cr and zero
mean. Hence, substituting P (7 ) into (4.4) and integrating would result in
/ 1 \ K/2f(vl«)= ( 2 ^ ) (det(Q))-'/"
x e x p ( - ^ { ¥ ; - â ) ^ Q “ 7 v - « ) ) . (4.5)
^In information theory, Gaussian distribution is a saddle point of many optimization problems. Therefore, we are safe to ignore the distributional uncertainties and considering the worst-case scenario, i.e., Gaussian case.
4.3. Optimal and Sub-Optimal Solutions 62
where det(-) denotes the déterminante operation and Q is the normalized covariance
matrix of the noise plus the estimation error, defined by
Q = ^ P { ( v + 7 ) (v + 7 )^}. (4.6)
Based on (4.5), it can be concluded that the optimal decision making procedure
should select from the set of the possible binary sequence 6 , which minimizes the
quadratic form
S { (p ,à )= { (p -à ) '^ (4.7)
For simplicity, assume that all the sub-bands experience equal estimation error,
hence, (4.7) can be expressed as
^ 2-2S (y, â) = \ \ c p - à f - i<p - â f , (4.8)
given that (f = ^ and â = ^ E&Li
Based on the above formulation, the optimal decision rule should compute S (y, a)
for all the possible 2^ possible binary sequences and further to search for the se
quence which returns the minimum value. However, this procedure is both compu
tational costly and timely, which makes this approach not a practical solution given
the demanding requirements for the spectrum sensing device, e.g., FCCs require
ments [4].
4.3.2 Optimal Solution For Fixed â
Let â to be fixed within a time frame, e.g, on average I sub-bands are occupied
within a time frame. Resulting in âi = l /K . Re-writing (4.8) based on the fixed âi
and minimizing leads to
S (y, Qi) = mm ||y - â ||" - - , (ÿ - S , f , (4.9)
where denotes the set of sequences with â = âj. It can be concluded from
(4.9) that the sequence ài which minimizes S {(p,oci) is the one consisting of I I ’s
corresponding to the largest I elements of (p. Therefore, the next task would be to
determine the I largest values of (p given that I is not known.
This can be done by ordering the elements of (p in terms of their magnitude. Let
4.3. Optimal and Sub-Optimal Solutions 63
fi be the K x K matrix which allows this transformation, hence we have u = flip,
where ui >U2, . . . , > u k - This would also result in /3 = f la . Now, the problem of
interest would be
S (u, ( 3 ) = min S (u. P i) . (4.10)
Using (4.8) and the definitions explained above we can re-write (4.10) as
I< 5 (u ,/3 i)= ^ (w fc - l)^
k=l^ 7 2 2 / I \ 2
It should be noted that ü = <p. More interestingly, (4.11) can be solved using the
recursion formula
5 (u, A ) = 5 (u, A _ i ) - 4 (ui + ' ~ ^ ^ | ) ~ 1 ) , (4.12)
with the initial condition
«5(u,/3o) = ^ 2 /^2 • (4-13)fe=l
Hence, the index Î which allows S < S { vl,P i) Wl ^ Î, determines P = Pi.
Consequently,
0, k > l
Finally, we have to re-store the original order of the indicator function, i.e., a =
f f^ p . As it can be observed by having a assuming a fixed â (which is acceptable
assumption in communication systems), we can reduce the search size from 2 ^
used in Section 4.3.1 to K , while maintaining the same performance. It should be
noted that f t f l ^ = f t ^ f l = I. Figure 4.1 shows the performance of the optimal
solution explained in this section for various SNR values and different I. As it can
be observed the proposed approach can easily detect the number of the occupied
sub-bands even in low SNR range. As S (u, /3) returns in a clear minimum value
at P = I. The effect of the SNR value of the performance can be easily observed,
i.e., as the SNR value decreases the S (u,/3) function will return a smoother curve
making the decision more prone to errors.
Even though the above procedure simplifies the ideal decision making process ex-
4.3. Optimal and Sub-Optimal Solutions 64
- - 1= 40, SNR = 0 dB — 1=40, SNR = 5 dB- ' 1= 40, SNR = 10 dB- -1= 70, SNR = 0 dB — 1 = 70, SNR = 5 dB ■-•1=70, SNR = 10dB
S ' 4 0
-10100
Figure 4.1: The performance evaluation of the optimal solution introduced in Section 4.3 for fixed â and K = 100.
plained in Section 4.3.1, it still needs to compute the recursive formula satisfying
(4.11). This would imply that the decision time would still be high, given sys
tems which employ high number of sub-bands, and motivates us to formulate a
sub-optimal solution to meet demanding spectrum sensing requirements.
4.3.3 Sub-optimal Solution
Consider the ideal case where the dimension of the observation sample approaches
infinity, i.e., limÜT —> oo, this results in
lim O' = £, and lim ip = 'tp,K ^ o o K -^ o o ^
(4.15)
where 'ip is the expected value of (p, which can be calculated based on the definition
in (4.1).
Furthermore, it should be noted that (4.8) can also be expressed as
K
s iv, 0 ) = ^ [<Pk - à k - { ( p - â ) Ÿ + ^2 '^ ^ 2 ( v - ô i f ■ (4.16)K a l
4.3. Optimal and Sub-Optimal Solutions 65
Hence, incorporating (4.15) into (4.16) will lead to
lim = . (4.17)K -> oo (TXj
Ignoring the constant term on the right hand side in (4.17) it can be concluded that
ÔC = 8 {p - {'ip - ^ ) ) , (4.18)
where £{ip) is an indicator function which returns 1 if ^ > 0 and 0 otherwise. It
can be observed from (4.18) that in order to obtain a satisfactory estimation of d,
an accurate knowledge of 'ip and should be provided or estimated. Therefore, the
performance of this solution is directly proportional to the number of sub-bands,
since the estimate of ip and ^ converges as K oo. In addition, employing ac
curate estimation approaches plays an important role in the performance of (4.18).
An example of an effective estimation method in this scenario would be minimum
variance linear estimation [86].
It would be important to evaluate the performance of the sub-optimal solution
explained above. The parameter which will reflect the accuracy of the proposed
approach would be the average probability of error, Pre- Assuming P(o; = 1) =
P (a = 0) = 0.5 by replacing (4.18) in (4.3)-(4.8) we have
/ 2 \ 0 .5
P r e = erfc ( ) , K = 1 (4.19)
where erfc(.) denotes the complementary error function and p represents the SNR
value.
It can be observed from above that the performance of the sub-optimal solution
is a function of two parameters, the number of sub-bands, K , and the SNR value.
More interestingly, the effect of (4.18), which can be modelled as high-pass filter,
will result in degradation of p approximately by a factor of {K — 1)/(AT -f p), i.e.
This degradation can be made arbitrarily small by choosing sufiiciently large K.
Figure 4.2 illustrates the effect of K on the performance of (4.18) for different SNR
4.4. Link to OFDMA based Systems 66
Optimal, SNR=0 dBOptimal, SNR=5 dB Optimal SNR=10 dB Sub-optimal, SNR=0 dB Sub-optimal, SNR=5 dB
- — ■ Sub-optimal, SNR= 10 dB
10 20 30 40 50 60 70 80 90 100
Figure 4.2: The performance comparison of the proposed optimal and sub-optimal solutions derived in Section 4.3.
values. It is shown in Figure 4.2 that as the SNR value decreases, the effect of K on
the performance decreases. Furthermore, the performance comparison of the sub-
optimal and optimal solution can be observed in Figure 4.2. Note that for increasing
values of SNR and K the performance gap of the sub-optimal and optimal solution
decreases, which was expected.
In the next section we will propose a novel joint sub-band level spectrum sensing
in OFDMA based environment, specifically for downlink scenarios. The physical
impairments existing in various wireless communication systems will be also consid
ered will be also considered in Section 4.4.
4.4 Link to OFDM A based System s
OFDMA is the most popular multi-carrier based transmission scheme for wide
band digital communication used in applications such as digital television and au
dio broadcasting and 40 mobile communications [87], due to its advantages, e.g.,
spectral and implementation efficiency and robustness towards propagation channel
fading. Hence, in this section, we will propose a novel joint sub-band level spectrum
sensing technique (inspired from the findings in Section 4.3).
4.4. Link to OFDMA based Systems 67
4.4.1 Transmitted Signal
The transmitted downlink OFDMA signal can be expressed in the matrix form,
Xfc = ^ X {T>{ak) X Sk), where Xfc is an J x 1 transmitted signal block, T>{ak)
denotes an M x M diagonal matrix with diagonal entries, ak acting as a frequency
mapping unit, indicating which sub-carriers are used for transmission, Sfc is an
M X 1 information-bearing symbol block with the covariance Og Im {M stands for
the number of subcarriers, and 1m for the identity matrix with the size M ), ^ is an
J X M ( J > M) tall pre-coding matrix with full column rank, and subscript k is the
block index. The pre-coding matrix Ÿ can be expressed as ^ . This
chapter is also focused on the CP based OFDMA systems. However, a solution forr -]T
zero padding based OFDMA system, i.e., where ^ 0 , is also provided
in Section 4.6.3.
4.4.2 Received Signal
In the absence of the signal, the spectrum sensing device can only receive noise,
otherwise, it receives a signal distorted by frequency-selective channel, timing offset
(denoted by e), frequency offset normalised by the subcarrier spacing (denoted by
e), and additive white Gaussian noise.
Please refer to Chapter 3 Section 1.2 for full details of received signal given the
mentioned distortions. Consider an J x 1 vector = [rfcj+i,rfcj+2v
then (3.3) can be formulated as the matrix form
rjt = Afc(A(ne)xfc -f V(ne)xfc_i) -f v&, (4.22)
where Ak = diag{A(^"^+^),
4.4.3 Join Sub-Band Level Spectrum Sensing in OFDMA
Systems
Incorporating the findings from Sections 4.3, a practical algorithm for sub-band
level sensing specifically for OFDM signals can be implemented using the following
steps: (the block diagram in Figure 4.3 demonstrates an overview of the proposed
algorithm)
S tep 1: Compute the BSD of the received signal.
4.4. Link to OFDMA based Systems 68
ReceivedSignal
Store Original
Figure 4.3: Flow chart of the proposed OFDMA sub-band level spectrum sensing.
This step consists of two operations, performing a DPT operation and a second
order average, respectively. Consider the signal model presented in (4.22), it can
be observed that due to the effect of CP the second term in the brackets, i.e.,
V(ne)xfc_i, is vanished. Hence, in order to obtain the BSD of the received signal
after the CP removal, = [rkM+i,^kM+2, - , an M-point DPT operation
is performed on r^, i.e., ffc =
Furthermore, at this stage, an M x M matrix, Cr, is computed by 5 " ' = ^k= oi^k^k)^
where N denotes the number of observation windows. In order not to converge the
presentation of the proposed technique consider the special case where, and e
are set to zero (Section 4.6 aims to overcome these offsets). Hence,
c ;= r (a) c l V M V h + WLiM, (4.23)
where P m — is an M x M diagonal matrix whose diagonal entries are
actually the channel frequency response, given that C m is an M x M circulant
channel matrix defined in [57].
Hence, the diagonal entry of Cr reads as
Qm — l^m| “b Afo 171 — 1, 2, • • • , M.L J m
(4.24)
Rem ark: A time domain window operation can be performed in order to reduce
the leakage between DPT bins. This leakage is due to the fact that DPT implicitly
assumes that the signal is periodic. Hence, when a discontinuity is observed in the
input signal it will spreads power all across the spectrum. One solution would be to
multiply the time series with a window function in the time domain before applying
the DPT [88].
4.4. Link to OFDMA based Systems 69
S tep 2 ; Order the BSD in ascending manner in terms of magnitude.
The values of BSD are sorted in increasing order, i.e., t 91 = where Ft is the
M x M sorting transformation matrix. The detailed explanation of the sorting
function can be found in Chapter 3 Section 1.3. The sorting function acts as a
weighting function to emphasise a particular region of the distribution function and
for different values of r the sort function allows us to focus on different regions of
the input distributions. This will have a direct effect on the error probability as it
will reduce the variance of the rth output i.e., t Qr considerably, see Section 4.5.
Furthermore, given a composite signal, by sorting the BSD in order of magnitude,
there would exist a point, namely the “knee-point”, where there is a sudden change
in the magnitude. This point will allow us to distinguish components from different
distributions. Hence, the next objective would be to identify the knee-point.
S tep 3: Apply differentiation on the ordered BSD.
The purpose of this step is to remove the energy component contained in all frequen
cies of interest. From (4.24) it can be observed that this would result in removing
the AWGN energy, A/q, from the BSD calculated in Step 1 and consequently allow
ing us to detect the knee point, k. In addition, the differentiation performed at
this stage will also allow us to exploit the frequency selectivity inherited in high
data rate communications [89]. This frequency selectivity will considerably have a
direct effect on the performance where all the sub-bands within the bandwidth of
the interest are all occupied.
In order to increase the reliability of the knee-point determination, a pre-determined
threshold, zu, can be employed. This will allow the proposed technique robust
towards heavily tailed noise and also power fluctuation due to limited sample size.
More interestingly, employing a threshold at this stage will signiflcantly improve
the performance of the proposed algorithm given the scenarios where the whole
bandwidth of interested is completely vacant. Thus, the knee-point will be the first
point in differentiate BSD where the corresponding magnitude is > zu.
Given the distribution of t q% one can determine the threshold value based on the
interference/throughput trade-off requirement of the system, see Section 4.5.3.
S tep 4: Zero force all the elements having a value less than the knee-point and
restore the original order.
Following the determination of the knee-point, the next step is to force the value of
all the elements in f <i' below the knee point to zero, i.e., t Qm = 0, Vm < «. Finally,
4.5. Performance Analysis for The Practical Solution 70
restore the original order of q using the inverse sorting transformation matrix used
in Step 2, i.e., q = t q ' (see Section 4.3.2). Sub-bands corresponding to a zero
value for m = 1, 2, • • • , M are assumed to be vacant.
From the above algorithm it can be observed that the proposed approach can be
used to offer spectrum sensing with sub-carrier level precision. This will benefit
system where low data rate/ bandwidth is required, e.g., sensor networks [90]. Since,
in most communications systems, operating devices require higher data rate, i.e.,
larger bandwidth, they would require more than one sub-carrier in order to transmit.
Hence, it would be to our interest to deliver a sub-band level accuracy.
4.5 Performance Analysis for The Practical Solu
tion
4.5.1 Asymptotic Probability of False Alarm and Threshold
In order to complete the construction of the spectrum sensing proposed in Section
4.4.3, we must provide a procedure to set the threshold vj. We propose to define
w in such a manner to keep PFA under a desired value. Given that for practical
applications, the PFA is often given a fixed value, such as 10% as per the FCCs
requirement. Hence, it would be necessary to derive the expression for PFA. As it
can be concluded from algorithm in Section 4.4.3, false alarm will occur when the
fluctuation resulting from two adjacent vacant sub-bands (i.e., noise only compo
nents) is greater than the threshold value. The exact probability density function
of the r th order statistic from any continues population is rather difficult to deal
with and in most cases requires numerical evaluation of a nontrivial integral [67].
Hence, in this section, we study the asymptotic behaviour of the order statistics of
q in order to determine a suitable threshold. Consider the special case of rig = 0.
It is understood that elements of q,„ (see (4.24)) given am = 0 follows [14]
/q (qrn|am=o) — X% (4.25)
where Xn denotes a central Chi squared distribution with N degrees of freedom.
Hence, after the sorting operation, t q (see 4.4.3 Step 2) based on the derivation
4.5. Performance Analysis for The Practical Solution 71
in Appendix E can be approximated as
t q . ~ g ( v n ■ — j (4.26)
where Q (a, b) denotes a Gaussian distribution with mean and variance equal to a
and b respectively, = f and $q(o) = i.e., CDF of qm|a,„=o- After
differentiation, f qrlar=ar-i=o will follow a Gaussian distribution with mean
Mq'r ~ (4.27)
and variance
+ j^o.5 • (4.28j
Hence, given a desired value of PFA in system, i.e., PFA< 'ijj the threshold value
can be obtained by
PFAr = q ( — (4.29)
where Q(.) is the standard Gaussian complementary CDF. As it can be observed,
the PFA is mainly dependent on two parameters, N and M. Since M is fixed
in any system, the performance can be improved by choosing longer observation
length, i.e., larger N. Furthermore, given a system with requirement of PFAr< w ,
the minimum threshold value, i.e., Wmm,r can be formulated as
^min,r — Mtqr (^r) • (4.30)
Since in the proposed algorithm one global threshold is used, zumin is set to max (tamin.r)
in order to meet the minimum requirement of the system. Given that in most sys
tems each sub-band has an equal PFA tolerance based on (4.27)-(4.30), rumin =
iUmin,K. In the case where the distribution of k is not known Wmin,K can be replaced
with Wrain,M to allow immunity in the worst case scenario.
4.5.2 Asymptotic Probability of “knee-point” Detection
As mentioned in Section 4.4.3, given a composite signal, by sorting the BSD in
order of magnitude, there would exist a point, namely the “fcnee-pomf’, where
4.5. Performance Analysis for The Practical Solution 72
there is a sudden change in the magnitude. In this section, we we will derive the
asymptotic distribution of the knee point. Given that the knee-point, «, is defined
as t q« = t Qk” t Q.k-1 where = 1 and 0:^-1 = 0. Based on Appendix E we have
t q . - i ~ S ( » " : ) , — j (4.31)
given that /g|a=o is defined in (4.25). Furthermore,
r , . ~ c (5.- (f--), " "<1: *’ '•1”") M
where [14]
9q (qm|am=l) ~ X n i^ P ) (4.33)
where, Xn (p) denote a non-central Chi squared distribution with the non-centrality
factor 2 p and H q (.) denoted the CDF of qm|am=i- Hence, after differentiation t q«
will follow a Gaussian distribution with mean
( i f ' " ) - i) (4-34)
and variance
fq;."' (M-K)OG
Given the distribution of a user activity within a region, i.e., /«(«;), probability of
detecting the right knee-point, PD«, can be formulated as
P D „ = / ° ° Q ( ” ~ / ^ ‘‘'-)/(K )dK .. (4.36)JQ \ /
11 I I. I y , —.1 II ✓
r(tu)
Given a system with requirement of PDk> r , the maximum threshold value, i.e.,
Wmax can be formulated as Wmax = F"^ (r). From the above derivations it can be
observed that the PD of the proposed sub-band level spectrum sensing is directly
prepositional to the number of sub-bands in the system of interest, M, observation
length, iV, the signal-to-noise ratio p at the sensing device and the threshold value,
VO.
4.5. Performance Analysis for The Practical Solution 73
4.5.3 Optimal Asymptotic Threshold Setting
It should be noted that value of the threshold, should be set in such a manner that
meets both the PD and PFA requirements of system, i.e., zumin < ro < zumax- More
specifically the exact value of threshold will be set in order to maximise the overall
throughput of the system while minimising the aggregate interference to the other
users. Given that the aggregate opportunistic throughput of the r th sub-band is
defined by
r..(ro) = J ^ 6 i( l-P F A j( ro ) ) (4.37)*eSr
where denotes the set of sub-carriers contained in the rth sub-band, bi denotes
the achievable rate of the i*^sub-carriers. Furthermore, the aggregate interference
to the user occupying the rth sub-band, can be evaluated as
^ Q (1 - PDi (a;)) (4.38)
where Q denotes the cost incurred if the user using the sub-carrier is inferred
with. As it can be observed from (4.37)-(4.38), opportunistic throughput and ag
gregate interference are directly promotional to PFA and PD respectively. Hence,
optimal threshold can be evaluated through receiver operating curve, known as
0.6
0.5
0.05 0.15
= 2, n , = 0 .5
= 5, rij = 0.5
= 10, = 0.5
_____N = 2. = 0.75
= 5, = 0.75
= 10, ti| = 0.75
0.25 0.3 0.35
Figure 4.4: ROC Curve of the proposed algorithm in Section 4.4.3 for p = 2 dB.
4.5. Performance Analysis for The Practical Solution 74
ROC curve. Figure 4.5.3 illustrates the ROC curve of the proposed algorithm in
Section 4.4.3. The ROC is shown for p = 2 dB, for different value of N and rjj^.
As it can be observed the performance improves for lower values of k . This was
expected as the noise power fluctuation for lower k would be less as it can be ob
served from (4.27)-(4.28) which would reflect in higher value for minra which in
turn would decrease the resulting PD. Detailed explanation on optimal threshold
setting based on interference/ throughput trade-off can be found in [50].
4.5.4 Computational complexity
The main complexity of the proposed scheme is due to the following stages:
1. The second-order time average: this stage requires x J complex multipli
cations and additions.
2. Discrete Fourier Transform: M —point DFT is implemented which introduces
the complexity by (9(Mlog(M)).
3. Sorting: the complexity of this stage on average is OMlog(M) (depending on
which sorting technique is employed).
4. Differentiation: this stage consists of subtracting every element of q from its
previous one, hence the computational complexity is given by 0{M ).
5. Knee point detection: This would add a further complexity of 0{M ).
6. Restoring the original order: a further complexity of O M log(M) is added due
to this stage.
Hence the overall computational complexity is:
JAT2 -K 3 0(M log(M )) 4- 2 0{M ). (4.39)
It can be concluded that the proposed algorithm is not very demanding in terms of
computational complexity.
4.6. Further Discussions 75
4.6 Further Discussions
4.6.1 Overcoming Timing Offset
In fact, the special case of rig = 0 can be hardly captured due to the lack of timing
synchronizations mechanism before the spectrum sensing component. Hence, in
order to handle the problem of unknown timing offset effectively, a “one ballot
ueto" policy is proposed to reject the vacancy of sub-bands. The policy is stated
as follows:
S tep 1; Form J x 1 vectors, = [rA=j+i+j,rA;j+2+j,...
where 6 denotes the offset in time.
S tep 2 : Compute according to (4.23), for ^ = 0, ( J — M), 2(J — M ), ..., M .
Apply the sub-band level spectrum sensing proposed in Section 4.4.3 on VS.
If for any value of Ô the energy within a sub-band falls above the knee-point value
it is understood that this particular sub-band is being used by an other user and is
not available for opportunistic use.
The key idea for proposing the “one ballot veto^’’ is that in the presence of a
signal, there exists such a S fulfilling the condition \n^ — S\ < J — M , and under
this condition, the proposed spectrum sensing scheme can successfully reject the
inter-block interference. In the absence of signal, is approximately constant
with respect to 5, due to constant energy of AWCN throughout the spectrum. Most
certainly, this stage will add to the overall complexity of the algorithm, however, in
order to increase the reliability of the sensing device, implementation of this stage
is necessary.
4.6.2 Overcoming the Energy Leakage
In all communication systems the transmitted signals are shaped by digital pulse-
shaping and analog band-pass filters according to a predefined frequency mask,
which in fact are not a perfect rectangular filters. As a result, the tail of the
transmitted signal in one sub-band might introduce significant interfering power
in adjacent sub-band provided that the signal is very strong. Moreover, the re
ceived samples at the sensing device are channelized with a non-ideal filter, which
also causes spectral leakage. Furthermore, the special case of perfect frequency
synchronisation, i.e., £ = 0 , can be hardly captured due to the lack of frequency
4.6. Further Discussions 76
synchronisation mechanism before the spectrum-sensing component. As a result
the energy of frequency bands will disperse [91] creating a challenge for sub-band
level spectrum sensing device. In order to combat this energy leakage and allow
robustness towards timing and frequency offsets requires the algorithm introduced
in Section 4.4.3 to be modified.
A possible solution would be to modify the condition stated in Step 4. Based on this
stage all the sub-carriers of a sub-band should be set to zero in order for that sub
band to be used by the opportunist user. Our proposed solution is to change this
condition by determining the vacancy of a sub-band not based on all, but at on a
portion of the sub-carriers within a sub-band. It is shown in [92] the energy leakage
has a significant effect on the sub-carriers situated at the edges of a sub-band while
the central sub-carriers are less effected. Based on this fact it can be concluded
that it is of our interest to determine the vacancy of a sub-band according to the
magnitude of the central sub-carriers. It was found through experiment that using
25 — 50% of the central sub-carriers we are able to overcome the energy leakage
problem significantly.
4.6.3 Extension to the ZP-based system
Let us start from the special case of = 0 and £ = 0. Using the result in [57],
we can easily justify that due to the implementation of ZP the second term in the
brackets, i.e., V(ne)xfc_i, of (4.22) is vanished. Therefore, performing a J-point
DFT on (4.22) yields
ffc = .FjAfc A(rie)xfc + .FjVfc, (4.40)
where JFj is an J x J DFT matrix normalized by the factor ( l ) / ( \ / j ) . Hence,
applying second order average on (4.40) will result in
(4.41)
where ^ C j is an J x J circulant channel matrix with T>j formed by
the corresponding channel frequency response. It is easy to observe that (4.41)
has the same form as (4.23). Therefore, the spectrum-sensing algorithm proposed
in Section 4.4.3 for the CP-based system can be straightforwardly applied on ZP-
based OFDMA systems. Furthermore, overcoming energy leakage policy explained
in Section 4.6.2 and the “one ballot veto” can also be employed in downlink ZP-
4.7. Simulation Results 77
Excess tap delay [ns] Relative power [dB]0 0
30 -170 -290 -3110 -8190 -17.2410 -20.8
Table 4.1: Extended Pedestrian A model
based OFDMA systems in order to show robustness towards physical impairments.
4.7 Simulation Results
Computer simulations were performed to evaluate the proposed spectrum-sensing
scheme. The system investigated in this section has M = 1,024 sub-carriers with the
sub-carrier spacing of 15 KHz,each frame consists of 7 OFDM blocks with the CP
length of J —M = 80, the sampling frequency is 30.72 MHz and the bandwidth is set
to 20 MHz. The channel is divided into 32 equal sub-bands, allowing each sub-band
to have 32 sub-carriers. The carrier frequency is also set at 5 GHz. We consider a
power delay profile, as given in Table 4.7 i.e., extended pedestrian A model, which
is defined for LTE based systems. This channel specified the frequency selective
channel between the transmitter and the spectrum sensing device. The sensing
device is assumed to be moving at the speed of 3 km/h. The SNR is defined by the
average received symbol energy to noise ratio at the sensing device.
Experiment 1
In this experiment we are aiming to observe the input behaviour and properties
at different stages of the algorithm explained in Section 4.4.3. Figure 4.5 shows
the step-by-step procedure which is carried out in the proposed solution. For this
experiment the SNR at the sensing device, p, is set to 2 dB and the observation
length N is equivalent to 2 frames duration. As it can be observed due to limited
sample size the estimated ESD at the sensing device experiences a large amount of
fluctuation. This energy fluctuation, due to limited size, will have a detrimental ef
fect on the performance of the existing spectrum sensing solution. More specifically,
this implies the conventional edge detection based approaches would not be deliv
ering an acceptable performance in such environments. The ESD is then ranked
4.7. Simulation Results 78
according to the magnitude, while the original ordering indexes are saved. As it can
be observed this stage will allow more robustness in terms of limited sample size as
it smoothens the fluctuation of the estimated ESD, while not effecting any statisti
cal properties of the input ESD. More interestingly, the existence of the knee-point
is very apparent, reflecting the point where the two components of the composite
signal, in this case noise and distorted transmitted signal, can be distinguished,
as mentioned in the Section 4.4.3. An other interesting observation which can be
made is that the differential values after the knee point, i.e., reflecting the occupied
sub-carriers, are higher comparing to the value denoting the noise only sub-carriers.
This is due to the frequency selective propagation channel and is explained in detail
in [79] [89]. This phenomenon will further improve the performance of the proposed
approach and reflects that we can use the frequency diversity of the received signal
to further verify the final results. An interesting point worth noting is the sudden
change which occurs at the tail of the differential ESD, this sudden taking place is
mainly due to the propagation channel. Hence, the tail of the differential ESD can
be employed for characterising and estimating the channel coefficients. This might
be an interesting factor when considering interference cancellation at the cognitive
device, however, it is out of the scope of this thesis.
n . n n n ,nn Estimated ESD of the Distorted Onginal Signal ESD
Ordered Estimated ESD
« 0.8 K nee Point
0.015
dioi0.005
0
0 0.2 0.4 0.6 0.8Index
1.2 1.4 1.6 1.8
x i o ’— Differentiated Ordered Estimated ESD| ! ' ' '
-K nee Point ---------- >
-
to L . !0 0.2 0.4 0.6 0.8
InJex1.2 1.4 1.6 1.8 2
xIO ^
Frequency
Figure 4.5: Step by step process of the proposed algorithm in Section 4.4.3 in order to determine the vacancy of unused sub-bands.
4.7. Simulation Results 79
Finally the frequency bands which reflect a zero value energy after the zero forcing
are considered as white space [16], i.e., vacant sub-bands and can be used via the
opportunistic spectrum access devices. The solution explained in Section 4.6.2 was
further employed to overcome the possible energy leakage in estimated ESD. Hence,
the flnal decision of whether a sub-band is vacant or not, was based on 50% of
the central sub-carriers in each sub-band. It is advisable to use > 25% of the
central sub-carriers to allow robustness in terms of the deep fading of the central
sub-carriers.
Experiment 2
The objective of this experiment is to evaluate the performance of the proposed
algorithm and compare it with the state-of-the-art sub-band level spectrum sensing
techniques. Firstly, we would like to compare the performance with the energy
detection based filter-bank approach, which is the most practical existing sub-band
level spectrum sensing solution. The energy detection is recognised as a blind
sensing scheme with advantages such as low complexity and low latency. However,
it is very sensitive to the noise uncertainty such that its performance is limited by
the SNR wall [9].
l o . e *s
1 — Pr opos ed algorithm N = 5 F ram es — Pr opos ed algorithm N = 10 F ram es
Energy D etection U=2 dB, N = 10 F ram es Energy D etection U=1 dB, N = 10 F ram es
1 1 i i i i0.01 0 .02 0 .03 0 .04 0.05 0.06 0 .07 0 .08 0 .09 0.1
Probability of F alse Alarm
Figure 4.6: ROC curve comparison of the proposed algorithm and energy detection based filter-bank approach.
The simulation performed for energy detection are based on noise uncertainty factor,
Î7 = 1,2 dB and the threshold is based on the assumed/ estimated noise power, while
4.7. Simulation Results 80
0 .8£
0.7£
0 .65
-6 - 4
) — () ^ * 6 > 0 6 --------
.........•] ........ ..................... / < : ..........
............
....... ................................... L '..: ............... .................
A / u 7 :/ ; f i / ' ;
—a — P roposed T echnique, N =2 F ram es — Pr opos ed Technique, N=5 F ram es —O — P roposed Technique, N =10 F ram es - O - Energy D etection, N= 10 F ram es, U= IdB “ O - Energy D etection, N= 10 F ram es, U= 2dB• ' K urtosis B ased D etection, N= lO F ram es• • E igenvalue b ase d D etectio, N=10 F ram es
i
f
i
0 2 4S N R (dB)
10
Figure 4.7: Performance comparison of the the proposed algorithm and energy detection based filter bank in terms of probability of detection for PFA = 0.01.
the real noise power varies with each Monte Carlo realisation by a certain degree.
Figure 4.6 illustrates the ROC curve of the proposed algorithm and the filter-bank
solution for SNR level of 3 dB for = 0.65. As it can be observed the proposed
algorithm can easily outperform the energy detection based filter-bank approach in
low SNR region. Furthermore, it can be observed that the performance of energy
detection is considerably dependent on the noise uncertainty factor. It is further
proved in [9] that increasing the observation length does not affect the performance
of the energy detection where the exact noise power is not known, i.e., U ^ 0 .
The performance of the proposed algorithm in terms of PD for various values of
observation length, N , is shown in Figure 4.7. The threshold is set in such a way to
ensure PFA< 0.01. As it can be observed the performance of the energy detection
based filter-bank drops dramatically as it hits the SNR wall while the proposed
algorithm degrades slowly for decreasing value of SNR. This can be particularly
useful in practical scenarios, reflecting the fact the proposed algorithm can detect
heavily shadowed signal. The performance comparison has been also carried out
for kurtosis based fllter-bank, where a decision is made based on the kurtosis test
which exploits the non-Gaussianity of communication signals [47] [48]. This scheme
features excellent accuracy at the price of large latency due to higher-order statistics.
However, as expected due to limited sample size, the proposed scheme can outper
form the kurtosis based detection by more than 10 dB. Filter-bank eigenvalue based
detection [35] was also used to carry out a more comprehensive performance com
4.8. Summary 81
parison of the proposed technique. The eigenvalue-based detection scheme exploits
orthogonality between the signal subspace and noise subspace using second order
stationarity features to offer highly reliable spectrum sensing [35]. For this perfor
mance comparison we used the maximum to minimum ratio of the eigenvalue as the
test statistic. It can be observed from Figure 4.7, the proposed technique outper
forms the eigenvalue based detection in low SNR environment. This was expected
since eigenvalue detection requires higher observation length.
4.8 Summary
In this chapter, a novel Bayesian model based approach for joint sub-band level
spectrum sensing has been proposed. This contributions is three-fold 1) re-defining
the objective of the sub-band level spectrum sensing device to a model estimator, 2)
deriving the optimal model selection estimator for sub-band level spectrum sensing
for fixed and variable number of users along with a sub-optimal solution based on
Bayesian statistical modelling and 3) proposing a practical model selection estimator
with relaxed sample size constraint and limited system knowledge for sub-band
spectrum sensing applications in Orthogonal Frequency-Division Multiple Access
and Non-Contiguous Orthogonal Frequency-Division Multiplexing systems. The
key idea behind the proposed approach is to exploit the second order frequency
diversity between signal and noise. Based on this approach after the ESD estimation
at the sensing device, they are ordered in terms of magnitude. The sorting operation
allows robust to noise power fluctuation due to limited observation length.
Furthermore, given a composite signal, by sorting the ESD in order of magnitude,
there would exist a point, namely the knee-point, where there is a sudden change
in the magnitude. This point will allow us to distinguish components from different
distributions. A differentiation stage is further employed. The objective of this
stage is two-fold, to remove the noise floor and to observe the knee-point. This
stage will also allow us to exploit the frequency selectivity inherited in high data rate
communications. This frequency selectivity will considerably have a direct effect on
the performance where the all the sub-bands within the bandwidth of the interest are
all occupied. In order to increase the reliability of the knee-point determination, a
pre-determined threshold is employed. This will allow the proposed technique robust
towards heavily tailed noise and also power fluctuation due to limited sample size.
More interestingly, employing a threshold at this stage will significantly improve
4.8. Summary 82
Approach Air-InterfaceInformation Synchronisation Computational
Complexity Latency
Filter Bank X X Low Low
Joint M ulti-Band Detection CompleteKnowledge / High High
Sequential M ulti-Band Detection CompleteKnowledge X High Low
W avelet Based Detection X X Medium Low
W igner-Ville Based Detection X / High High
Bayesian Based Appraoch PartialKnowledge X Medium Low
Table 4.2: Comparison of the state-of-the-art local sub-band spectrum sensing approaches and proposed Bayesian approach.
the performance of the proposed algorithm given the scenarios where the whole
bandwidth of interested is completely vacant. Thus, the knee-point will be the first
point in differentiate ESD where the corresponding magnitude is greater than the
predetermined threshold. Following the determination of the knee-point, the next
step is to force the value of all the elements in the ordered differentiated ESD below
the knee point to zero. Finally, the ordered differentiated ESD is re-sorted using
the inverse sorting transformation matrix used in the sorting stage. The sub-bands
corresponding to zero value ESD are considered to be white space and are available
for opportunistic spectrum access use.
More interestingly, the proposed sub-band level spectrum sensing does not require
a priori knowledge of noise power and the propagation channel gain, and is designed
in such a way to show robustness towards energy leakage, without introducing high
complexity to the overall system, i.e., less the cubic computational complexity.
The proposed sub-band level spectrum sensing approach is analytically evaluated
through asymptotic probability of false alarm and probability of detection. It was
proved that the performance of the proposed approach is directly promotional to
the observation length, the total number of the sub-carriers and the SNR of the
received signal at the sensing device. Computer simulations are carried out in
order to evaluate the effectiveness of the proposed scheme. Through appropriate
threshold adjustments, the proposed approach can be extended to state- of-the-art
OFDMA/NC-OFDM joint sub-band level spectrum sensing.
hapter 5Conclusion and Future Work
5.1 Conclusion
In this thesis the original problem of local spectrum sensing in cellular networks was
investigated. Firstly, a comprehensive review of the existing narrow-band spectrum
sensing algorithms was presented. The existing approaches were divided into three
main categories:
• Explo iting energy difference. The most well-known spectrum sensing
approach developed under this category is the energy detection [11]. The
energy detection is recognised as a blind sensing scheme with advantages such
as low complexity and low latency. However, it is very sensitive to the noise
uncertainty, resulting in that its performance that is limited by the SNR
wall [9].
• Explo iting s ta tio n a rity difference. The initial works of spectrum sensing
through stationarity difference can be traced back to work of Dandawate et al
in [25]. This category of spectrum sensing schemes can trade latency with high
sensing reliability. Furthermore, they are sensitive to the noise uncertainty.
• Explo iting th e d istr ib u tio n difference. Given that in almost all commu
nication system models, noise is assumed to be additive white and Gaussian,
one can determine the vacancy of a particular frequency band by observing
the difference of the received signals distribution and that of the AWGN.
These schemes feature excellent accuracy at the price of large latency due to
83
5.1. Conclusion 84
higher-order statistics. A critical point of such sensing approaches is that their
performance degrades significantly when signals are approximately Gaussian.
Furthermore, the problem of sub-band level spectrum sensing was explored and it
was found that the literature of sub-band spectrum sensing is rather limited at this
time. However, a full survey of the state-of-the-art sub-band level spectrum sensing
was also provided.
In terms of contribution firstly, a novel differential energy detection scheme for multi
carrier systems, namely, cluster-based differential energy detection, which can form
fast and reliable decisions on spectrum availability even in very low SNR environ
ment, has been proposed. It has several distinctive features including low latency,
high accuracy reasonable and computational complexity, as well as robustness to
low SNR. For example, the proposed scheme can reach 90% in probability of detec
tion and 10% in probability of false alarm for an SNR as low as —21 dB, while the
observation window is equivalent to 2 multi-carrier symbol duration. The proposed
scheme at this stage is specially designed for sensing multi-carrier sources but we
would argue since most of the current and future mobile networks are multi-carrier
based systems, the proposed scheme has potentially wide-ranging practical appfi-
cations. The proposed approach can deliver desirable performance in high density
communication networks and urban area due to its ability to allow robustness in low
SNR environments. Furthermore, the clustered-based differential energy detection
can be employed in vehicular communication systems.
The key idea of the proposed scheme is to exploit the channel frequency diversity
inherent in high data-rate communications using the clustered differential ordered
energy spectral density. Initially, the ESD of the received signal is estimated. Specif
ically, after the ESD computation, the clustering operation is utilized to group un
correlated subcarriers based on the coherence bandwidth to enjoy a good frequency
diversity. The knowledge of coherence bandwidth does not need to be very accurate
(we employ the reciprocal of the maximal channel delay). Furthermore, making use
of order statistics of the estimated ESD, we further increase the reliability of the
sensing algorithm. This will allow us to smooth the fluctuation in noise ESD, re
sulted from limited observation length with effecting the statistics of the received
signal. Hence, this will stage will have a direct effect on the PFA of the proposed
approach.
In order to exploit the second order moment diversity of the observed signal, a
differential operation is performed on the rank ordered ESD. When the channel is
5.1. Conclusion 85
frequency selective and the noise is white, the differential process can effectively re
move the noise floor resulting in elimination of the noise uncertainty impact which
is the main factor making energy detection reluctant [9]. Furthermore, the differ
ential stage will allow us to exploit the frequency selectivity which available in the
received signal. At the final stage of the proposed scheme, the differential rank or
dered ESD within different clusters are linearly combined in order to further reduce
the effect of impulse/spike noise. Binary hypothesis testing is then applied on either
the maximum or the extremal quotient (maximum-to-minimum ratio) depending on
the wireless channel characteristics of the sensed environment. More importantly,
the proposed spectrum sensing scheme is designed to allow robustness in terms of
both, time and frequency offset, without compromising computational complexity.
To analytically evaluate the proposed scheme, both PD and PFA were derived for
Rayleigh fading channels. The closed-form expression showed a clear relationship
between the sensing performance and the cluster size, i.e., channel coherence band
width, which is an indicator of the diversity gain. Computer simulations are carried
out in order to evaluate the effectiveness of the proposed approach and to compare
the performance of the proposed scheme with state-of-the-art spectrum sensing
schemes where up to 10 dB gain in performance can be observed. This would imply
that employing the proposed approach in a communication system will make the
network of interest more robust to hidden node problem, i.e., mitigate interference
to heavily shadowed licensed users. More interestingly, it has been shown through
simulation that the proposed spectrum sensing algorithm has a long convergence
time, which allows it a suitable in delay sensitive systems.
Secondly, a novel Bayesian model based approach for joint sub-band level spectrum
sensing has been proposed. This contributions is three-fold 1) re-defining the objec
tive of the sub-band level spectrum sensing device to a model estimator, 2) deriving
the optimal model selection estimator for sub-band level spectrum sensing for fixed
and variable number of users along with a sub-optimal solution based on Bayesian
statistical modelling and 3) proposing a practical model selection estimator with re
laxed sample size constraint and limited system knowledge for sub-band spectrum
sensing applications in Orthogonal Frequency-Division Multiple Access and Non-
Contiguous Orthogonal Frequency-Division Multiplexing based systems. The key
idea behind the proposed approach is to exploit the second order frequency diversity
between signal and noise. Based on this approach, after the ESD estimation at the
sensing device, the ESD is ordered in terms of magnitude. The sorting operation
5.1. Conclusion 86
allows robust to noise power fluctuation due to limited observation length.
Furthermore, given a composite signal, by sorting the ESD in order of magnitude,
there would exist a point, namely the knee-point, where there is a sudden change
in the magnitude. This point will allow us to distinguish components from different
distributions. A differentiation stage is further employed. The objective of this
stage is two-fold, to remove the noise floor and to observe the knee-point. This
stage will also allow us to exploit the frequency selectivity inherited in high data
rate communications. This frequency selectivity will considerably have a direct
effect on the performance where the all the sub-bands within the bandwidth of
the interest are all occupied. In order to increase the rehability of the knee-point
determination, a pre-determined threshold is employed. This will ensure robustness
of the proposed technique against heavily tailed noise and also power fluctuation due
to limited sample size. More interestingly, employing a threshold at this stage will
significantly improve the performance of the proposed algorithm given the scenarios
where the whole bandwidth of interested is completely vacant. Thus, the knee-point
will be the first point in differentiate ESD where the corresponding magnitude
is greater than the predetermined threshold. Following the determination of the
knee-point, the next step is to force the value of all the elements in the ordered
differentiated ESD below the knee point to zero. Finally, the ordered differentiated
ESD is re-sorted using the inverse sorting transformation matrix used in the sorting
stage. The sub-bands corresponding to zero value ESD are considered to be white
space and hence available for opportunistic spectrum access use.
More interestingly, the proposed sub-band level spectrum sensing does not require
a priori knowledge of noise power and the propagation channel gain, and is designed
in such a way to show robustness towards energy leakage, without introducing high
complexity to the overall system, i.e., less the cubic computational complexity.
The proposed sub-band level spectrum sensing approach is analytically evaluated
through asymptotic probability of false alarm and probability of detection. It was
proved that the performance of the proposed approach is directly proportional to
the observation length, the total number of the sub-carriers and the SNR of the
received signal at the sensing device. Computer simulations are carried out in
order to evaluate the effectiveness of the proposed scheme. Through appropriate
threshold adjustments, the proposed approach can be extended to state- of-the-art
OFDMA/NC-OFDM joint sub-band level spectrum sensing.
5.2. Future Work 87
5.2 Future Work
To this point, further work in the context of local spectrum sensing can be consid
ered. These works can either be extensions to the proposed schemes or they can be
new and are summarised as follows:
Firstly, the clustered differential energy detection based approach proposed in Chap
ter 3, assumes a priori knowledge about some of the air-interface parameters , e.g.
number of sub-carriers and channel propagation bandwidth. It would be interest
ing if this work can be extended to a blind spectrum sensing technique where no
information of the operating users are required.
Secondly, the proposed sub-band level spectrum sensing technique presented in
Chapter 4, is developed for downlink OFDMA scenarios. Hence, this calls for an
updated system model which also accommodates uplink scenarios. It should be
noted that the problem of uplink spectrum sensing is a very promising research
topic, since, to this date very limited research has been carried on in this context.
This topic can be very challenging since it has to account for multi-dimensional
synchronisations due to multiple users. Furthermore, it would be desirable to extend
the proposed sub-band level spectrum sensing technique so that it would operate
without a prior knowledge of the signalling information.
Thirdly, for making the most out of the opportunities, spectrum sensing device
should keep track of variations in spectrum availability and should make predictions.
Hence, the history of the spectrum usage information can be used for predicting the
future prole of the spectrum. It will very interesting to incorporate the knowledge
about currently active users or prediction of the spectrum usage in to the final
decision made by the spectrum sensing device.
Appendix A
Probability Distribution Function of
the Differential ESD
The moment generating function (MGF) of the output of the differential process in
step S2 in Section 4.4.3, 9^1% , (which is expressed as the product of the MGF of
two central Chi squared random variables shown in (3.23)) is given by [24]
A4(g |9Yo) = A4(gi+i|%) x Ai(%|%o)
= (1 - 2 t) - ^ X (1 + 2 t) - ^ = (1 - 4 f ) " ^ (A.l)
where A4 (a) denotes the MGF of random variable a, and t denotes the time-domain
index. Hence, (A.l) indicates that follows a summation of K i.i.d. Laplace
distributions (with location parameter fi — 0 and scaling parameter ^ = 2), whose
their MGF functions are given by [24]
M { ^ ) = _ ^2 2
where J f denotes a random variable following a Laplace distribution. Moreover, we
know for a fact that qhlHo is non-negative due to the sorting process. Therefore, it
can be concluded that distribution of can be further simplified to summation
of K exponential distributions with scale factor 0.5. Hence, the probability density
function (p.d.f.) of
(A.3)
Appendix B Distribution of the Test I statistics
Since q ' is i.i.d., for 1 < m < L, we would have
FAax(ql%o)(Ai) = Pr(max(q'|?4o) < Ai)
= Pr(9z, > A i,çi, > 9m Vm 7 L^Hq)
UPr(9l_i > A i,9 i > ^ Vm f L - 1 |% ) U . . .
UPr(9 i > A i , q[ > 9 ^ Vm f l|74o)
= P r % > A i , 9 ^ > q'rrym f L\Hq)
^ Ai, 9 > q '^ m ^ L — l \ H o ) + . . .
+Pr(9i > Ai,9l > ôLVm f 1|% )
= [FgilWo(Ai)]^- (B.l)
PFA for Test I can then be evaluated using the cumulative density function (CDF)
of max(q'|'Ho)j given qi\Ho variables are i.i.d.
Test I : PFA= Pr(max(q'|74o) > Ai)
= 1 - [Fmax(q'|Wo) (Ai)]V---------- '
âPr(m ax(q'|H o)<Ai)
= 1 - ( A i ) ] . ( B .2 )
89
90
Furthermore, the PD for Test I can be expressed as
Test I : PD = Pr(max(Ay|94i) > Ai)
= 1 - [FmaxCqOI?^! ( A i ) ]L
= i - n (B.3)1 = 1 -------1-PDTi
Appendix C Distribution of the Test II statistics
The Hypothesis for Test II is based on the ratio of maximum to minimum of q^|% ,
whose CDF can be computed using
P - / ' max(q'IHo) ^ ,
= P r ( ^ > A2,5i > r„.Vm Lkq'i < %'ig + l|Wo)U
• . . U P r ( g - > A 2 , î1 > ÿm Vm ^ L & ÿ i < % | % )
= P r ( ^ > A z .îi > %,Vm + < % !% )+
• •. + P r ( ^ > A2,ÿ i > fLVm f 2) & g i < (C.l)
Due to max / min(q'|?4o) > 0, we can further express (C.l) as
fCX)poo pA 2 P
■ F n iax /m in (q '|?^ o ) ( ^ 2 ) — I I { o c , P ) d ( XJo J (x=0
poo pX zP
+ / / fq'L-iQil'^o (^» ^) do: d;d + . . .Jo J q := 0
pOO pX20C+ / /g' g '170 (a,/3)d/3 do: (C.2)
Jo J b= o
Once iTiax/min(q'|7{o) (^2)18 obtained, the PFA for Test II can be easily computed
by
Test II : PFA = 1 - F ’m a x /m in (q '|W o )(A 2 )- (C.3)
Furthermore, for Text II we have
Test II : PD = Pr(m ax/m in(q'|'H i) > Ag)
~ t “ [Fmax(g'/9'l'^i) (A2 )]
(D= 1 - n F(A72/A7d^0i?ii (A2) (C.4)
1- p o p
91
Appendix D
Probability Distribution Function of
ESD Over Rayleigh Fading Channel
Using the recursion for (3.36), the following result is obtained
Q niax, h)
= g - ” Tj^/2-i{axy) + QK-i{ax,h)
K - l
= Y ] e " In{axy)Qi{ax,b). (D.l)n = 0
Applying (D.l) into (3.40), the integration part can be computed by
/>oo
= g ( ^ y rn = 0
rOO/ rr^^^~^^exp~^ ^^Qi(ax,P)dx (D.2)
J x = 0
where p = 2, and = y/X. Using combination of Bessel functions and
exponentials in [73], the above terms can be evaluated.
92
Appendix E
Asymptotic Distribution of The rth
Order Statistics
We will first approximate the rth order statistics of the uniform distribution and
using probability-integral transformation transformation we derive the r th order
statistics from any continues distribution. Given that Ur is the rth ordered value
from uniform distribution with a total of M elements, we have
= (E.1)
Replacing Zr = in above leads to
/Zr (z) = (E-2)
where
© = (r - 1) log ^1 -t- -h(M - r ) log ■ (E.3)
93
94
Using Taylor series expansion
0 =z
2
— (r — 1) — { M — r )
n 1-/LI
L/
/T 2 (T 2— (r — 1) — ------ (M — r)w I - u
cr 3(M — r)
(E.4)
Given M o o and r / M —)■ 77 we can make the following approximations
M — r + 1
r { M + 2)
0.5
M r ] J0.5
M(1 — 77)
0.5
(E.6)
(EG)1 -M [ { M - r + l ) { M + 2 ) _
Substituting the above in to (E.4) it can be found that the coefficient of z and z ^ / 3
tends to 0 as 71 0 0 while the coefiicient of —z^ / 2 converges to 1 for increasing
value of n . Using these findings and replacing them into (E.4) and ignoring terms
of order n ~ ^ ' ^ and higher we have, lim - cx) 0 = — Using Stirling’s formula, i.e.,
k \ = \/27rexp { — k ) and the approximation
^ (M + 1 )^ (M + 2)0-5y .r -0 .5 ( M - r + l ) ^ - r + 0 5
(E.7)
( M + 1 ) M + o.5
we obtain the following
(E.8)
Hence,
(E.9)
The probability-integral transformation allows us to conclude that the same asymp
totic distribution holds for any continuous distribution as long as the appropriate
] i and a are replaced. Furthermore it has been shown in [67] the asymptotic mean
and standard deviation of the rth value of the random variable x containing total
elements of M can be approximated as
/ = Fx (?7) &: (T =(77(1 - 77))°-^(/x (//))
MO-5 (E.IO)
Bibliography
[1] Q. Zhao and B. M. Sadler, “A survey of dynamic spectrum access”, IEEE
Signal Process. Mag., vol. 24, no. 3, pp. 79 -89, May. 2007.
[2] J. Mitola III and G. Q. Maguire, “Cognitive radio: making software radios
more personal”, IEEE Per s. Commun., vol. 6, no. 4, pp. 13 -18, Aug. 1999.
[3] A. Ghasemi and E. S. Sousa, “Spectrum sensing in cognitive radio networks:
requirements, challenges and design trade-offs”, IEEE Commun. Mag., vol. 46,
no. 4, pp. 32 -39, Apr. 2008.
[4] Federal Communications Commission, “Notice of proposed rule making and
order: Facilitating opportunities for flexible, efficient, and reliable spectrum
use employing cognitive radio technologies”, pp. 03-108, Feb. 2005.
[5] “IEEE draft standard for information technology telecommunications and in
formation exchange between systems local and metropolitan area networks
specific requirements part 22.1: Standard to enhance harmful interference pro
tection for low power licensed devices operating in tv broadcast bands”, IEEE
P802.22.1/D8, pp. 1 -151, Aug. 2010.
[6] V. Goncalves and S. Pollin, “The value of sensing for TV white spaces”, in
IEEE DySPAN, Aachen, Germany, pp. 231 -241, May 2011.
[7] Ofcom, “Implementing geolocation [online]
http: / / stakeholders.ofcom.org.uk/consultations / geolocation / statement”,
Sep. 2011.
[8] R. S. Rappaport, Wireless Communications: Principles and Practice, Number
4th. Prentice Hall, 2001.
95
Bibliography 96
[9] R. Tandra and A. Sahai, “SNR walls for signal detection”, IEEE J. of Sel.
Topics Signal Process., vol. 2, no. 1, pp. 4 -17, Feb. 2008.
[10] G. Zhao, J. Ma, G. Y. Li, T. Wu, Y. Kwon, A. Soong, and C. Yang, “Spatial
spectrum holes for cognitive radio with relay-assisted directional transmission”,
IEEE Trans, on Wireless Commun.,, vol. 8, no. 10, pp. 5270 -5279, 2009.
[11] H. Urkowitz, “Energy detection of unknown deterministic signals” , Proc. IEEE,
vol. 55, no. 4, pp. 523 - 531, Apr. 1967.
[12] A. Sonnenschein and P. M. Fishman, “Radiometric detection of spread-
spectrum signals in noise of uncertain power” , IEEE Transactions on Aerospace
and Electronic Systems, vol. 28, no. 3, pp. 654 -660, Jul. 1992.
[13] N.S. Shankar, C. Cordeiro, and K. Challapali, “Spectrum agile radios: uti
lization and sensing architectures”, IEEE DySPAN, Maryland, USA, pp. 160
-169, 2005.
[14] F. F. Digham, M. S. Alouini, and M. K. Simon, “On the energy detection of
unknown signals over fading channels”, IEEE Trans. Commun., vol. 55, no. 1,
pp. 21 -24, 2007.
[15] P. Qihang, Z. Kun, W. Jun, and L. Shaoqian, ” , IEEE PIMRC,Helsinki,
Finland.
[16] S. Haykin, “Cognitive radio: brain-empowered wireless communications”,
IEEE J. Sel. Areas in Commun., vol. 23, no. 2, pp. 201 - 220, Feb. 2005.
[17] R. Durrett, Probability: Theory and Examples, Number 3rd. Belmont,CA:
Duxbury, 2004.
[18] J. Wu, T. Luo, and G. Yue, “An energy detection algorithm based on double
threshold in cognitive radio systems”, in IEEE ICISE, Nanjing, China, pp.
493 -496, Dec. 2009.
[19] Y. M. Kim, G. Zheng, S. Hwan Sohn, and J. M. Kim, “An alternative energy
detection using sliding window for cognitive radio system”, in IEEE ICACT,
Phoenix Park, Korea, pp. 481-485, Feb. 2008.
[20] K. Kim, Y. Xin, and S. Rangarajan, “Energy detection based spectrum sensing
for cognitive radio: An experimental study”, in IEEE GLOBECOM, Florida,
USA, pp. 1-5, Dec. 2010.
Bibliography 97
[21] M. P. Olivieri, G. Barnett, A. Lackpour, A. Davis, and P. Ngo, “A scalable
dynamic spectrum allocation system with interference mitigation for teams of
spectrally agile software defined radios” , IEEE DySPAN 2005, Maryland, USA,
pp. 170 -179, 2005.
[22] S. Lai and A. Mishra, “A look ahead scheme for adaptive spectrum utilization”,
RAWCON ’03. Proceedings Radio and Wireless Conference, 2003., pp. 83 - 86,
2003.
[23] F. Weidling, D. Datla, V. Petty, P. Krishnan, and G.J. Minden, “A framework
for r.f. spectrum measurements and analysis”, IEEE DySPAN, Maryland, USA,
pp. 573 -576, 2005.
[24] A. Papoulis and S. U. Pillai, Probability, Random Variables and Stochastic
Processes, Number 4th. McGraw-Hill New York, 2002.
[25] A. V. Dandawate and G. B. Giannakis, “Statistical tests for presence of cyclo-
stationarity”, IEEE Trans. Signal Process., vol. 42, no. 9, pp. 2355 -2369, Sep.
1994.
[26] S. M. Cabric, D. Mishra and R. W. Brodersen, “Implementation issues in
spectrum sensing for cognitive radios”, Conference Record of the Thirty-Eighth
Asilomar on Signals, Systems and Computers., vol. 1, pp. 772 - 776 Vol.l, Nov.
2004.
[27] N. Khambekar, L. Dong, and V. Chaudhary, “Utilizing OFDM guard inter
val for spectrum sensing”, IEEE Wireless Communications and Networking
Conference, WCNC 2007, pp. 38 -42, Mar. 2007.
[28] K. Maeda, A. Benjebbour, T. Asai, T. Furuno, and T. Ohya, “Recognition
among ofdm-based systems utilizing cyclostationarity-inducing transmission”,
IEEE DySPAN, pp. 516 -523, 2007.
[29] P. D. Sutton, K. E. Nolan, and L. E. Doyle, “Cyclostationary signatures for
rendezvous in ofdm-based dynamic spectrum access networks”, IEEE DySPAN,
Dublin, Ireland, pp. 220 -231, 2007.
[30] P. D. Sutton, J. Lotze, K. E. Nolan, and L. E. Doyle, “Cyclostationary signa
ture detection in multipath rayleigh fading environments” , IEEE CrownCom,
pp. 408 -413, 2007.
Bibliography 98
[31] J. Lunden, V. Koivunen, A. Huttunen, and H. V. Poor, “Spectrum sensing in
cognitive radios based on multiple cyclic frequencies”, CrownCom 2007., 2007.
[32] S. Chaudhari, V. Koivunen, and H. V. Poor, “Autocorrelation-based decentral
ized sequential detection of OFDM signals in cognitive radios”, IEEE Trans.
Signal Process., vol. 57, no. 7, pp. 2690 -2700, Jul. 2009.
[33] Y. Zeng and Y.C Liang, “Covariance based signal detections for cognitive
radio”, IEEE DySPAN, Dublin, Ireland, April 2007.
[34] J. Ma, G. Y. Li, and B. H. Juang, “Signal processing in cognitive radio”, Proc.
IEEE, vol. 97, no. 5, pp. 805 -823, May 2009.
[35] Y. Zeng and Y. C. Liang, “Eigenvalue-based spectrum sensing algorithms for
cognitive radio”, IEEE Trans. Commun., vol. 57, no. 6, pp. 1784 -1793, Jun.
2009.
[36] P. Wang, J. Fang, N. Han, and H. Li, “Multiantenna-assisted spectrum sensing
for cognitive radio”, IEEE Trans. Veh. Technol, vol. 59, no. 4, pp. 1791 -1800,
May 2010.
[37] Y. Zeng and Y. C. Liang, “Maximum-minimum eigenvalue detection for cog
nitive radio”, IEEE PIMRC, Athens, Greece, pp. 1 -5, 2007.
[38] Y. Zeng, C. L. Koh, and Y. C. Liang, “Maximum eigenvalue detection: Theory
and application”, IEEE ICC, pp. 4160 -4164, May 2008.
[39] H. S. Chen, W. Gao, and D. G. Daut, “Signature based spectrum sensing
algorithms for IEEE 802.22 WRAN”, IEEE ICC, 2007.
[40] D. Cabric, S. M. Mishra, and R. W. Brodersen, “Implementation issues in
spectrum sensing for cognitive radios”, Conference Record of the Thirty-Eighth
Asilomar Conference on Signals, Systems and Computers, vol. 1, pp. 772 - 776
Vol.l, 2004.
[41] H. S. Chen, W. Gao, and D. Daut, “Spectrum sensing for OFDM systems
employing pilot tones”, IEEE Trans. Wireless Commun., vol. 8 , no. 12, pp.
5862 -5870, Dec. 2009.
[42] H. Gao, S. Daoud, A. Wilzeck, and T. Kaiser, “Practical issues in spectrum
sensing for multi-carrier system employing pilot tones” . Applied Sciences in
Biomedical and Communication Technologies (ISABEL), pp. 1-5, Nov. 2010.
Bibliography 99
[43] C. Cordeiro, M. Ghosh, D. Cavalcanti, and K. Challapali, “Spectrum sensing
for dynamic spectrum access of tv bands”, IEEE CrownCom,, pp. 225 -233,
Aug. 2007.
[44] T. M. Cover and J. A. Thomas, Elements of Information Theory, Number 2nd.
Wiley-Interscience, 2006.
[45] Y. L. Zhang, Q. Y. Zhang, and T. Melodia, “A frequency-domain entropy-
based detector for robust spectrum sensing in cognitive radio networks”, IEEE
Commun. Lett., vol. 14, no. 6, pp. 533 -535, Jun. 2010.
[46] Y. Zhang, Q. Zhang, and S. Wu, “Entropy-based robust spectrum sensing in
cognitive radio”, lE T Communications, vol. 4, no. 4, pp. 428 -436, May 2010.
[47] D. Denkovski, V. Atanasovski, and L. Gavrilovska, “Efficient mid-end spec
trum sensing implementation for cognitive radio applications based on USRP2
devices”, in IEEE COCORA, Budapest, Hungary, Apr. 2011.
[48] D. Denkovski, V. Atanasovski, and L. Gavrilovska, “HOS based goodness-of-fit
testing signal detection”, IEEE Commun. Lett, pp. 1 -4, To appear 2012.
[49] B. Farhang-Boroujeny, “Filter bank spectrum sensing for cognitive radios”.
Signal Processing, IEEE Transactions on, vol. 56, no. 5, pp. 1801 -1811, May.
2008.
[50] Z. Quan, S. Cui, A. H. Sayed, and H. V. Poor, “Optimal multiband joint de
tection for spectrum sensing in cognitive radio networks”, IEEE Trans. Signal
Process., vol. 57, no. 3, pp. 1128 -1140, Mar. 2009.
[51] S. J. Kim, G. Li, and G.B. Giannakis, “Multi-band cognitive radio spectrum
sensing for quality-of-service traffic”, IEEE Trans. Wireless Commun., vol. 10,
no. 10, pp. 3506 -3515, oct. 2011.
[52] Z. Govindarajulu, Sequential Statistical Procedures, Academic Press, Univer
sity of Michigan, 1975.
[53] S. Mallat and W. L. Hwang, “Singularity detection and processing with
wavelets”, IEEE Trans. Inf. Theory, vol. 38, no. 2, pp. 617-643, Mar. 1992.
[54] Z. Tian and G. B. Giannakis, “A wavelet approach to wideband spectrum sens
ing for cognitive radios”, International Conference on CrownCom Proceedings,
CrownCom, Jun. 2006.
Bibliography 100
[55] M. Biagi, V. Polli, and J. A. A. Freitas, “An image processing approach to
distributed access for multiantenna cognitive radios”, in IEEE ISWCS, York,
UK, pp. 621 -625, Sep. 2010.
[56] B. Boashash, Time-Prequency Signal Analysis: Methods and Applications,
John Wiley & Sons, New York, 1992.
[57] Z. Wang and G. B. Giannakis, “Wireless multicarrier communications”, IEEE
Signal Process. Mag., vol. 17, no. 3, pp. 29 -48, May 2000.
[58] T. Keller and L. Hanzo, OFDM and MC-CDMA: A Primer, Wiley - IEEE,
2007.
[59] M. Movahhedian, Y. Ma, and R. Tafazolli, “Blind CFO estimation for linearly
precoded OFDMA uplink”, IEEE Trans. Signal Process., vol. 58, no. 9, pp.
4698 - 4710, Sep. 2010.
[60] M. L. Gao, “The self-adapt spectrum management formula base on cluster
model”, in IEEE ICMT, Zurich, Switzerland, pp. 901 -903, Jul. 2011.
[61] A. B. McDonald and T. F. Znati, “A mobility-based framework for adaptive
clustering in wireless ad hoc networks”, IEEE J. Sel. Areas in Commun., vol.
17, no. 8, pp. 1466 -1487, Aug. 1999.
[62] E. Dall’Anese and G. B. Giannakis, “Distributed cognitive spectrum sensing
via group sparse total least-squares”, in IEEE CAMSAP, San Juan, Puerto
Rico, pp. 341 -344, Dec. 2011.
[63] C. Ragusa, A. Liotta, and G. Pavlou, “An adaptive clustering approach for
the management of dynamic systems” , IEEE J. Sel. Areas in Commun., vol.
23, no. 12, pp. 2223 - 2235, Dec. 2005.
[64] R. C. Hardie and K. E. Earner, “Rank conditioned rank selection filters for
signal restoration”, IEEE Trans. Image Process., vol. 3, no. 2, pp. 192 -206,
Mar. 1994.
[65] R.C. Hardie and C. Boncelet, “LUM filters: a class of rank-order-based filters
for smoothing and sharpening”, IEEE Trans. Signal Process., vol. 41, no. 3,
pp. 1061 -1076, Mar. 1993.
[66] J. H. Lin, T. M. Sellke, and E. J. Coyle, “Adaptive stack filtering under
the mean absolute error criterion”, IEEE Trans. Acoust, Speech and Signal
Process., vol. 38, no. 6, pp. 938 -954, Jun. 1990.
Bibliography 101
[67] H. A. David and H. N. Nagaraja, Order statistics, John Wiley and sons
Hoboken. New Jersey, 2003.
[68] R. Sedgewick and K. Wayne, Algorithms, Number 4th. Addison-Wesley Pro
fessional, 2011.
[69] J. Serra, Image Analysis and Mathematical Morphology: Theoretical Advances,
vol. 2, New York: Academic, 1988.
[70] S. R. Sternberg, “Biological image processing”. Computer, vol. 16, no. 1, pp.
22-34, Jan. 1983.
[71] A. Taherpour, M. Nasiri-Kenari, and S. Gazor, “Multiple antenna spectrum
sensing in cognitive radios” , IEEE Trans. Wireless Commun., vol. 9, no. 2, pp.
814 -823, Feb. 2010.
[72] H. R. Karimi, “Geolocation databases for white space devices in the UHF
TV bands: Specification of maximum permitted emission levels” , in IEEE
DySPAN, Aachen, Germany, pp. 443 -454, May 2011.
[73] I. S. Gradshteyn and I. M. Ryzhik, Table of Integrals, Series, and Products,
Academic Press, 2000.
[74] F. F. Digham, M. S. Alouini, and M. K. Simon, “On the energy detection of
unknown signals over fading channels” , IEEE Trans. Commun., vol. 55, no. 1,
pp. 21 -24, Jan. 2007.
[75] A. H. Nuttall, “Some integrals involving the QM-function” , Naval Underwater
System Center (NU SC ) Technical Report, May 1974.
[76] 3GPP TS 25.814, “Physical layer aspects for evolved universal terrestrial radio
access” , vol. Rel. 7, 2007.
[77] Final Report on Link Level and System Level Channel Models, “IST-2003-
507581 WINNER D 5.4”, vol. V 1.4, no. Available: http://www.ist-winner.org,
Nov. 2005.
[78] Z. Lu, Y. Ma, and R. Tafazolli, “A first-order cyclostationarity based energy
detection approach for non-cooperative spectrum sensing” , in IEEE PIMRC,
Istanbul, Turkey, pp. 554 -559, Sep. 2010.
[79] P. Cheraghi, Y. Ma, and R. Tafazolli, “Frequency-domain differential energy
detection based on extreme statistics for OFDM source sensing”, in IEEE
VTC, Budapest, Hungary, pp. 1 -5, May 2011.
Bibliography 102
[80] D. N. Lawley, “Tests of significance for the latent roots of covariance and
correlation matrices”, Biometrika, vol. 43, pp. 128-136, Jun. 1956.
[81] M. S. Bartlett, “A note on the multiplying factor for various approxima
tions”, J. R. Stat Sac., vol. 16, pp. 296-298, 1954.
[82] H. Akaike, “A new look at the statistical model identification”, IEEE Trans.
Autom. Control, vol. 19, no. 6, pp. 716 - 723, Dec. 1974.
[83] G. Schwarz, “Estimating the dimension of a model”. The Annals of Statistics,
vol. 6, no. 2, pp. 461-464, Mar. 1978.
[84] M. Wax and T. Kailath, “Detection of signals by information theoretic criteria”,
IEEE Trans. Acoust., Speech, Signal Process., vol. 33, no. 2, pp. 387 - 392, Apr.
1985.
[85] T. Bayes, “An essay toward solving a problem in the doctrine of chances”,
Phil. Trans. Roy. Soc., vol. 53, pp. 370-418, 1763.
[86] Y. Theodor and U. Shaked, “Robust discrete-time minimum-variance filtering”,
IEEE Trans. Signal Process., vol. 44, no. 2, pp. 181 -189, feb. 1996.
[87] A. Larmo, M. Lindstrom, M. Meyer, G. Pelletier, J. Torsner, and H. Wiemann,
“The LTE link-layer design”, IEEE Commun. Magazine, vol. 47, no. 4, pp. 52
-59, apr. 2009.
[88] F.J. Harris, “On the use of windows for harmonic analysis with the discrete
fourier transform” , Proc. IEEE, vol. 66, no. 1, pp. 51 - 83, Jan. 1978.
[89] P. Cheraghi, Y. Ma, R. Tafazolli, and Z. Lu, “A novel low complexity differ
ential energy detection for sensing ofdm sources in low SNR environment”, in
IEEE GLOBECOM, Texas, USA, Dec. 2011.
[90] EU-ICT EXALTED Project [Online] http://www.ictexalted.eu, ” .
[91] J. Liu and J. Li, “Parameter estimation and error reduction for ofdm-based
WLANs”, IEEE Trans. Mobile Comput, vol. 3, no. 2, pp. 152 - 163, Jun. 2004.
[92] T. H. Yu, O. Sekkat, S. Rodriguez-Parera, D. Markovic, and D. Cabric, “A
wideband spectrum-sensing processor with adaptive detection threshold and
sensing time” , IEEE Trans. Circuits Syst. I, Reg. Papers, vol. PP, no. 99, pp.
1, 2011.
Reproduced with permission of copyright owner. Further reproduction prohibited without permission.