optimal pulses robust to insufficient synchronization for
TRANSCRIPT
Optimal Pulses Robust to Insufficient Synchronization for
OFDM/OQAM Systems
Yu Zhao1, Xihong Chen
1, and Qun Zhang
2
1Air and Missile Defense College of Air Force Engineering University, Xi’an 710051, China
2Information and Navigation College of Air Force Engineering University, Xi’an 710077, China
Email: {sunzy54321, xhchen0315217, pepper24}@163.com
Abstract—In this paper, a novel pulse shaping method is
presented for Orthogonal frequency division multiplexing with
offset quadrature amplitude modulation (OFDM/OQAM)
systems. Designed as a linear combination of normalized
Hermite functions, the proposed pulse is optimized with respect
to interference caused by insufficient synchronization such as
carrier frequency offset (CFO) and timing offset (TO). The
optimization problem, based on signal-to-interference ratio (SIR)
maximization criterion, is carried out using an efficient iterative
algorithm. Simulation results show that this optimal pulse
outperforms traditional prototype pulses in the presence of
insufficient synchronization. Index Terms—OFDM/OQAM, pulse shaping, SIR, carrier
frequency offset, timing offset
I. INTRODUCTION
OFDM is employed in various systems due to its
robustness against time dispersion caused by multipath
propagation. These systems are lattice structures formed
by time-frequency shifts of a rectangular prototype pulse.
However, the cyclic prefix or guard time interval used to
prevent inter-symbol interference (ISI) will cost a loss of
spectral efficiency. Moreover, the rectangular pulse
which exhibits a sinc-shaped frequency response will
result in serious inter-carrier interference (ICI) in
frequency dispersion channels. In order to counteract the
above drawbacks, some pulse-shaping OFDM systems
(PS-OFDM) were proposed [1]-[3]. With lower sidelobe
and faster spectral decay pulse shape, PS-OFDM systems
are more robust to frequency dispersion caused by
Doppler effect and carrier frequency offset (CFO).
However, the underlying requirement in PS-OFDM is
that the pulses must be orthogonal or biorthogonal to
each other. However, a well localized orthogonal pulses
set cannot be achieved for unit density structures due to
Balian-Low theorem. Thus, many structures employ half
lattice density which results in a loss of bandwidth
efficiency.
Filter Bank Multicarrier (FBMC) is intensively studied
now as an alternative to OFDM in order to increase
bandwidth efficiency and to improve the robustness
Manuscript received December 10, 2014; revised May 12, 2015.
This work was supported by National Natural Science Fund of China
under Grant no. 60971100.
Corresponding author email: [email protected].
against various dispersive channels [4]. The most popular
FBMC technology is the OFDM/OQAM modulation
which has already been introduced in the TIA’s Digital
Radio Technical Standards and considered in WRAN
(IEEE802.22) [5]. In OFDM/OQAM systems, the
modulation used for each subcarrier is a staggered offset
QAM denoted as OQAM. The OQAM method introduces
a half symbol duration time offset between the real and
imaginary components of QAM (quadrature amplitude
modulation) constellation. With this approach, the
orthogonality constraint only holds in real field.
Therefore, a well time-frequency localized pulse can be
achieved in systems with unit lattice density.
In this paper, a novel OFDM/OQAM pulse shape
modeled as a linear combination of normalized Hermite
Journal of Communications Vol. 10, No. 5, May 2015
301©2015 Journal of Communications
doi:10.12720/jcm.10.5.301-307
OFDM/OQAM systems suffer from high sensitivity to
some desynchronization like CFO and timing offset (TO)
[6], [7]. CFO is mainly caused by different local
oscillators of the transmitter and receiver. TO may occur
when the actual sampling instant at the receiver is
different from the optimal one. Normally, CFO and TO
estimation are done prior to the demodulation, but the
estimation is usually not perfect. Fortunately, the
prototype pulses used in the systems are able to provide
different immunity against CFO and TO [8]. It has been
shown that well frequency-localized pulse, such as square
root raised cosine (SRRC), is in favor for the TO
tolerance. If equally severe CFO and TO are present,
isotropically time-frequency localized pulses such as
IOTA and weighted Hermite pulse (WHP) [9], [10], are
more suited. However, none of them can remain the best
place for all the cases. Therefore, design of robust pulses
to insufficient synchronization is an attractive research
topic. In [11], pulse shapes that have minimum ICI power
for a given value of CFO in PS-OFDM systems are
derived. In [12], a pulse shape design procedure which is
optimal with respect to interference due to CFO in
OFDM/OQAM systems is introduced. Recently, a novel
optimized generalized Gaussian pulse (OGGP) is
proposed [13]. The pulse has lower side lobes and faster
decay in time domain compared to conventional pulse
shaping methods. In [14], the author formulated a
problem of direct optimization of the filter coefficients to
minimize the stopband energy. Therefore, the optimized
filter (DOFC) can achieve significantly good frequency
selectivity.
functions is proposed. The pulse is designed to be robust
to insufficient synchronization. Firstly, the exact SIR
expression of the demodulated symbol versus CFO and
TO is derived. Then, the coefficients of Hermite
functions are optimized based on SIR maximization
criterion. Finally, the optimization problem is simplified
to a series of quadratically constrained quadratic
programs (QCQP) using an iterative algorithm. By
semidefinite relaxation, the QCQP can be translated to
semidefinite programming (SDP) problems which are
easy to be solved. Theoretical analysis and simulation
results show that the optimal pulse outperforms
traditional pulses especially in the systems with large
CFO and TO.
The rest of this paper is organized as follows. The
system model of OFDM/OQAM in the presence of CFO
and TO is described in the next section. In Section 3,
design of robust pulse is formulated as a SIR
maximization problem and the optimization algorithm is
presented. Then performance of the optimal pulse is
tested by numerical simulations in Section 4. Finally
some conclusions are drawn in Section 5.
II. SYSTEM MODEL IN THE PRESENCE OF CFO AND TO
A. Baseband Model for OFDM/OQAM
The transmitted baseband signal in the OFDM/OQAM
system is generated as:
0
12
, 0
0
, ( )
( ) ( )mn
Mj j mF t
m n
m n
m n t
s t a e e g t n
g
(1)
where ( / 2)( )mn m n is the phase term resulting
from the staggering rule of OQAM; ,m na denotes the
real-valued symbol located on the mth subcarrier at
time instant 0n . ,2m ka and ,2 1m ka can be obtained
by taking the real and imaginary components of a
complex-valued symbol from QAM constellation
respectively. They are assumed to be independent and
identically distributed with zero mean and average energy2
a . 0F is the inter-carrier spacing,
0 is the symbol
duration time (0 0 1/ 2F ); M is the subcarrier number
which is supposed to be even. ( )g t denotes the
symmetrical real-valued prototype pulse in this paper. Assuming an ideal channel, complete reconstruction of
the symbol ,m na can be obtained according to the
following orthogonality condition:
( )( )
, ,
0 0
( ), ( ) {( )
(( ) , ( ) )}
( , )
n l m k n l m k
m n k l R
g
g t g t j
A k m l n F
m k n l
(2)
where ( , ) is the 2-D Kronecker delta function.
( , )gA is the ambiguity function expressed as:
* 2( , ) ( / 2) ( / 2) j t
gR
A g t g t e dt (3)
Hence, the orthogonality can be ensured by finding a
real-valued and even prototype pulse of which the
ambiguity function satisfies [5]:
0 0(2 ,2 ) ( , )gA u vF u v , ( , )u v (4)
B. Effects of CFO and TO
In our analysis, phase offset and channel noise is
disregarded. Considering the insufficient synchronization,
we introduce a carrier frequency offset f and a timing
offset t .
At the receiver, the demodulated signal for
time-frequency position 0 0( , )n m is performed by
projecting the received signal onto the set of functions
0 0,{ ( )}m ng t with a modified inner product. Then the
demodulated data symbol can be expressed as:
0 0 0 0
0 0
2
, ,
2
,
ˆ ( ), ( )
{ ( ) ( ) }
j ft
m n m nR
j ft
m nR
a e s t g t t
e s t g t t dt
(5)
where {} is the real part operator. Substituting (1) and
(3) in (5) and introducing 0m m k and
0n n l ,
after some steps of straightforward manipulations, one
obtains that:
31 2 4
0 0 0 0
1
, ,
0
0 0
ˆ {
( , )}
Mjj j j
m n m k n l
m n
g
a a e e e e
A l t kF f
(6)
where 1 0 0 0 0(2 2 )m F t n f t f ,
2 0kn ,
3 ( )2
k l kl
, 4 0 0( )kF t l f . Then, it is
easy to separate the demodulated data symbol into the
useful and interference part, i.e.:
0 0 0 0 0 0 0 0, , , ,ˆ { }m n m n m n m na a J (7)0
with
1
0 0, ( , )j
m n ge A t f (8)
and
31 2 4
0 0 0 0, ,
0 0
0 0( , )
jj j j
m n m k n l
k l
g
J e a e e e
A l t kF f
(9)
In this paper, the phase estimation of the desired signal
section is assumed to be perfect at the receiver. Then, a
simple one-tap zero-forcing (ZF) equalizer is used to
compensate the phase rotation. Finally, it leads to the
equalized signal expressed as:
0 0 0 0
0 0 0 0
0 0 0 0
0 0
, ,
, ,
, ,
,
ˆ { } { }
( , )
m n m nZF
m n m n
m n m n
g m n ZF
Ja a
H H
A t f a J
(10)
Journal of Communications Vol. 10, No. 5, May 2015
302©2015 Journal of Communications
where 1
0 0
j
m nH e
is the ZF equalizer coefficient at the
0 0( , )m n subcarrier. Noting that as long as the prototype
pulse is an even real-valued function, gA is also real
and even. Concerning the first part of the second equality
of (10), the desired symbol is attenuated by ( , )gA t f
which is caused by CFO and TO. As 2 0kn ,
2 1j
e . Therefore, the interference power can be
written as:
3 4
2
( )
0 0
0 0
2 2
3 4 0 0
0 0
2
4 0 0
2
0(
2
2
2
0, ),
{ ( , )}
cos ( ) ( , )
(0.5 0.5cos 2 ) ( , )
J ZF
j
g
k l
g
k l
g
k lt f
a
a
a
k l
P E J
e A l t kF f
A l t kF f
A l t kF f
(11)
Then, the SIR expression writes:
2
2
, 0 0
0 0
( , )
(( ,
, ) ( ))
,
g
k l g
k l
SA t f
It f A l t kF f
R t f
(12)
III. DESIGN OF ROBUST PULSE
A. Pulse Shape Design
Equation (12) illustrates that the SIR depends on the
prototype pulse through its ambiguity function while
CFO and TO are confirmed. The pulses are usually
designed based on the following criterion: 1) energy
concentration, 2) rapid-decay, 3) spectrum-nulling and 4)
channel characteristics. In the wireless communication,
SIR is the predominant parameter that characterizes the
system performance. Therefore, we search for the robust
prototype pulse based on SIR maximization criterion, i.e.:
( )( ) argmaxopt g tg t SIR (13)
The optimal pulse is designed as a linear combination
of Hermite functions which are maximally localized both
in time and frequency. The nth order Hermite function
( )nh t is defined by:
21/4( ) 2 2 ! 2n t
n nh t n H t e (14)
where ( )nH t is the nth order Hermite polynomial with
0 ( ) 1H t , 1( ) 2H t t , and it can be computed
recursively by: 1 1( ) 2 ( ) 2 ( ) n n nH t tH t nH t . Noting
that Hermite function corresponding to 0n is the
Gaussian pulse and it provides the best time and
frequency localization. Hermite function is the
eigenvector of its Fourier transformation:
( ) ( )n n nF h t h t , 2 jn
n e is the corresponding
eigenvalue. For 2 , 0,1,2...n i i , both ( )nh t and its
Fourier transformation are real-valued even functions.
Then the proposed pulse is expressed as: 1
2
0
( ) ( )L
i i
i
g t h t
(15)
In (15), the L most concentrated Hermite functions are
chosen. The coefficients i are assumed to be real.
Substituting (15) in (3), we obtain:
1 1
0 0
( , ) ( , )L L
g p q pq
p q
A A
(16)
where* 2
2 2( , ) ( / 2) ( / 2) j t
pq p qR
A h t h t e dt ,and
they can be calculated according to [15]. Substituting (16)
in (12) and introducing ' '( , ) ( , ) ( , )pq p qA v A v A v , we
obtain:
' '
' '
' '
' '
' '
0 , , , 1
' '
0 , , , 1
( , , , )
( , , , )
p q p qp q p q L
p q p qp q p q L
S p q p q
SIRI p q p q
(17)
' '( , , , ) ( , )AS p fq p tq (18)
, 0
0
0
' '
0
( , ) (( , , , ), )k l
k lI p q p t f A t kF fq l
(19)
Once well localized pulses are used, it is sufficient to
consider the energy perturbation only from neighboring
symbols with negligible performance loss. Thus, we only
consider 1k and 1l in (19). Finally, the
optimization task amounts to finding , 0,1,... 1i i L
so that the objective function ( , )SIR t f is maximized.
B. Optimization Problem Formulation
A new variable w a a is introduced to simply the
SIR expression. denotes matrix Kronecker product
and 0 2 1...T
L a . Then, as a function of w , the
SIR expression (17) can be reorganized in the following
matrix form:
T
TSIR
w Sw
w Iw (20)
where ( ) ijsS , ( )ijinI , 20 , 1i j L are symmetrical
and positive definite (SPD) matrix respectively defined
by: ' '( , , , )ijs S p q p q ,
' '( , , , )ijin p q p qI , i pL q ,
' 'j p L q .
The ambiguity function can also be reorganized in a
matrix form:
( , ) ( , ) T
gA a A a (21)
where ( , ) ( ) ijaA , 0 , 1 i j L is defined by
( , ) ij ija A . In [10], it was proposed that the
orthogonality constraints (4) are applied at the origin
( , ) 0u v and other L-1 nearest grid points due to the
good localization property of ( )g t . Then the coefficients
1 2 1... T
La can be determined by substituting (21)
in (4) and solving the resulting L equations.
Journal of Communications Vol. 10, No. 5, May 2015
303©2015 Journal of Communications
As ( , ) gA is real-valued, the orthogonality conditions
(4) can be transformed to:
2
0 0(2 ,2 ) ( , )gA u vF u v (22)
Substituting (16) in (22), we obtain:
( , ) ( , )T u vv uw B w (23)
where ( , ) ( ) iju v bB ,20 1, Lji , is defined by:
0 0(2 ,2 )ij ub A vF , i pL q , ' 'j p L q . Considering
the effect of insufficient synchronization, the constraints
on the nulls of the ambiguity function is not necessary
[16]. Moreover, for the sake of design freedom, we relax
the constraints (23) to: (0,0) 1Tw B w .
Besides the orthogonality condition constraints on w ,
there are some other inherent constraints due to
Kronecker product operator. In fact, two kinds of equality
constraints and one inequality constraint are introduced
(see Appendix): 0Cw ; 0T
ijw D w , 20 , 1 i j L ;
0Gw . Eventually the optimization problem can be
formulated as shown as follows:
(P): maxT
Tw
w Sw
w Iw
s.t. 1: (0,0 1)TC w B w , 2 : 0T
ijC w D w
3: 0C Cw , 4 : 0C Gw (24)
C. Iterative Algorithm for SIR Maximization
The optimization problem (P) is a fractional
programming problem, and the objective function is
non-convex in this case. In general, such problem is
NP-hard. To solve this problem, it will be convenient to
adopt Dinkelbach’s parametric approach as follows [17].
We define as the set of feasible solutions of problem
(P). Associated with this problem, the equivalent
parametric problem is:
( ) max[ ( , ) ]
T TF fw
w w Sw w Iw (25)
( ) arg max ( , )
fw
w w (26)
If there exists 0 for which ( ) 0 F ,
( ) w w is an optimal solution of (P) [17]. Then, an
effective iterative algorithm (known as Dinkelbach’s
method) is summarized as follows:
Input: outerL (maximum number of iterations)
0outer (convergence tolerance)
1: Calculate the largest eigenvalue and its
corresponding eigenvector of 1
I S . If ,
terminate( is an globally optimal solution);
Otherwise go to Step 2
2: 0i , 0 0
3: do whileouteri L and
1( , )
i i outerf w
4: 1i i
5: Solve 1( )iF
to obtain the optimal
solution
iw (inner loop)
6: T T
i i i i iw Sw w Iw
7: end do
8: return
It has been proved that convergence to the optimal
solution is guaranteed. Thus, problem (P) is transformed
to a series of QCQP problems:
1max ( )T
i w
w S I w , s.t. 1C , 2C , 3C , 4C (27)
Due to the variety of , optimization problem in (27)
is a non-convex quadratic program. Semidefinite
relaxation (SDR) method can be used to generate the
suboptimal solutions with quite encouraging results [18].
A crucial step for SDR is to observe that
Tr( ( ))T Tw Hw H ww , where Tr( ) stands for the
trace of a matrix. We introduce a variable TM ww ,
and note that it is equivalent to M being a rank one and
positive semidefinite (PSD) matrix. Thus, we obtain the
following equivalent formulation of (27):
1max Tr(( ) )i M
S I M
s.t. 1: Tr( (0,0) ) 1C B M
2 : Tr( ) 0ijC D M , 20 , 1i j L
3: Tr( ) 0TC C CM , 4 :C Gw 0
5: 0, ( ) 1C rank M M (28)
here, 0M indicates that
that except for the non-convex rank-one constraint
( ) 1rank M in 5C , the remaining problem is convex.
Thus, by dropping the rank-one constraint, we obtain a
relaxed semidefinite programming (SDP) problem which
can be efficiently solved using the general purpose CVX
package [19].
IV. NUMERICAL RESULTS
Among the prototype pulses that have been considered
for our studies in this paper, the IOTA and WHP design
have isotropic time-frequency localization, and OGGP
has the lowest time-domain side lobes. The DOFC design,
on the other hand, has the lowest side lobes in frequency
domain. In this section, the SIR performance of the
optimal pulse will be compared with that of the above
prototype pulses. The experiments are established
through a 20MHz channel where M=256 subcarriers and
4-QAM are employed. Rectangular time-frequency
lattice is set to 0 0 1 2F . f and t are
normalized to 0F and
0 , respectively.
In Fig. 1, it is showed that system performance in
terms of SIR is greatly improved when increasing the
number L of Hermite functions. Nevertheless, the SIR
Journal of Communications Vol. 10, No. 5, May 2015
304©2015 Journal of
is PSD it can be seen M
improvement decreases with the number L. Moreover,
increasing L will induce computational complexity and
even cause instability in the optimization process.
Therefore, we choose L=8 as a compromise according to
the simulation.
Fig. 1. SIR comparison versus the number of Hermite functions
Fig. 2. SIR comparison versus CFO with TO=0
Fig. 3. SIR comparison versus TO with CFO=0
The curves for SIR versus normalized CFO are
presented in Fig. 2 for the case where timing is perfect. In
the comparison, the optimization was redone for each set
of f . The figure shows that the optimal pulse
significantly outperforms other prototype pulses for f
larger than 0.04 at the cost of the SIR of nearly 25 dB at
0f . It can be also concluded that more SIR
performance gain can be obtained for larger CFO.
The curves for SIR versus normalized TO are
presented in Fig .
As is shown in the figure, the optimal pulse has the
highest SIR for t larger than 0.025. We also note that,
the optimal pulse performs slightly inferior to other
pulses in the small TO regions. However, in this case, the
SIR level is still higher than 20 dB. Similar results can be
concluded that more SIR performance gain can be
obtained for larger TO.
Fig. 4. SIR comparison versus CFO and TO
As is shown in Fig. 2 and Fig. 3, in small CFO or TO
regions, the optimal pulse performs slightly inferior to
other pulses. This is because the constraints on the nulls
of the ambiguity function of the proposed pulse are
abandoned. It can also be seen that in the presence of
CFO error, the SIR curves of OGGP and DOFC are
displayed in the reversed order compared to TO case. It
confirms that an improvement of the robustness against
CFO error can be addressed by using the pulse like
OGGP which has good time localization. On the other
hand, good time localization means bad frequency
localization, therefore, it is not preferable in the TO error.
Fig. 4 gives the SIR comparison where the same
increment step for CFO and TO is set. WHP and IOTA
pulses lead to approximately the same performance, so
IOTA is not included in this comparison. As is shown,
the WHP design performances better than OGGP and
DOFC due to the fact that WHP provide identical
localization in time and frequency. Moreover, it can be
seen that the optimal pulse is superior to other pulses. It
confirms once again the efficiency of our method to
acquire a prototype pulse robust to insufficient
synchronization.
V. CONCLUSION
In this paper, we have proposed a novel approach to
design prototype pulses that lead to robust performance
2 4 6 8 10 120
5
10
15
20
L
SIR
(dB
)
CFO=TO=0.05
CFO=TO=0.1
CFO=TO=0.15
CFO=TO=0.2
0 0.05 0.1 0.15 0.2 0.25 0.30
5
10
15
20
25
30
normalized CFO
SIR
(dB
)
optimal pulse
OGGP
WHP
IOTA
DOFC
0 0.05 0.1 0.15 0.25
10
15
20
25
30
normalized TO
SIR
(dB
)
optimal pulse
OGGP
WHP
IOTA
DOFC
0.05 0.1 0.150
5
10
15
normalized CFO=TO
SIR
(dB
)
optimal pulse
OGGP
WHP
DOFC
Journal of Communications Vol. 10, No. 5, May 2015
305©2015 Journal of Communications
. 3 for the case where CFO is fixed to zero
against insufficient synchronization for OFDM/OQAM
systems. The robust pulse was modeled as a linear
combination of normalized Hermite functions. Exact SIR
expression of demodulated symbol versus CFO and TO is
derived. The pulse shaping approach based on SIR
maximization criterion has been formulated as a
non-convex fractional programming problem. An
efficient iterative algorithm was proposed to transform
the NP-hard problem to a series of QCQP problems
which can be efficiently solved using SDR method.
Numerical simulations demonstrated that the optimal
pulse design method provides a significant gain over
traditional pulse shaping method. The advantage is
especially prominent in the case of systems with large
carrier frequency and timing offset.
APPENDIX
From the expression of w , we note that:
(i) ( 1) 1i L j jL iw w , 0 2,0i L j i ;
(ii)2
1 ( 1) 1iL j iL i j L jw w w , 0 2, 2i L i j L ;
(iii)2
2 0iL i iw , 0 1i L .
For (i), this equality constraint can be summarized by:
0Cw , where C is a 2 2L L matrix defined by
0 2
TT T T
L C C C 0 . [ ]iC is a
2( 1)i L matrix
given by:
1 ( 1)
1 1
0
i mn
n i L m
n mL i
otherwise
C (29)
For (ii), we deduce this kind of equality constraints:
0T
ij w D w , where ijD is a 2 2L L matrix defined by:
2 2( ( 1))
TT T T
ij ijiL L L i L L
D 0 E 0 , ijE is a 2L L matrix
given by:
1 1, 1
0.5 , ( 1) 1[ ]
0.5 ( 1) 1,
0
ij mn
m j n iL j
m i n j L j
m j L j n i
otherwise
E (30)
For (iii), we deduce this inequality constraint:
0Gw , where G is a 2 2L L matrix defined by:
1 ( 1),0 1
0mn
n m L m L
otherwise
G (31)
ACKNOWLEDGMENT
This work was supported by National Natural Science
Fund of China under Grant no. 60971100.
REFERENCES
[1] D. Schafhuber, G. Matz, and F. Hlawatsch, “Pulse-shaping
OFDM/BFDM systems for time-varying channels: ISI/ICI analysis,
optimal pulse design, and efficient implementation,” in Proc. IEEE
PIMRC, vol. 3, Lisbon, Portugal, 2002, pp. 1012-1016.
[2] K. Liu, T. Kadous, and A. M. Sayeed, “Orthogonal time-frequency
signaling over doubly dispersive channels,” IEEE Trans. on
Information Theory, vol. 50, no. 11, pp. 2583-2603, 2004.
[3] T. Strohmer and S. Beaver, “Optimal OFDM design for
Time-Frequency dispersive channels,” IEEE Trans. on
Communications, vol. 51, no. 7, pp. 1111-1122, 2003.
[4] B. Farhang-Boroujeny, “OFDM versus filter bank multicarrier,”
IEEE Signal Processing Magazine, vol. 28, no. 3, pp. 92-112,
2011.
[5] J. F. Du and S. Signell, “Classic OFDM systems and pulse-shaping
OFDM/OQAM systems,” Technical Report (KTH-Royal Institute
of Technology), pp. 1-32, 2007.
[6] H. Saeedi-Sourck, Y. Wu, J. W. M. Bergmans, S. Sadri, and B.
Farhang-Boroujeny, “Sensitivity analysis of offset QAM
multicarrier systems to residual carrier frequency and timing
offsets,” Signal Processing, vol. 91, pp. 1604-1612, 2011.
[7] H. Lin, M. Gharba, and P. Siohan, “Impact of time and carrier
frequency offsets on the FBMC/OQAM modulation scheme,”
Signal Processing, vol. 102, pp. 151-162, 2014.
[8] A. Sahin, I. Guvenc, and H. Arslan, “A survey on multicarrier
communications: Prototype filters, lattice structures, and
implementation aspects,” IEEE Communications Surveys &
Tutorials, vol. 16, no. 3, pp. 1312-1338, 2014.
[9] P. Siohan, C. Siclet, and N. Lacaille, “Analysis and design of
OFDM/OQAM systems based on flterbank theory,” IEEE Trans.
on Signal Processing, vol. 50, no. 5, pp. 1170-1183, 2002.
[10] R. Haas and J. C. Belfiore, “A time-frequency well-localized pulse
for multiple carrier transmission,” Wireless Personal
Communications, vol. 5, pp. 1-18, 1997.
[11] G. Lin, L. Lundheim, and N. Holte, “Optimal pulses robust to
carrier frequency offset for OFDM/QAM systems,” IEEE
Communications Letters, vol. 12, no. 3, pp. 161-163, 2008.
[12] G. Lin, N. Holte, and L. Lundheim, “Design of robust pulses to
carrier frequency offset for OFDM/OQAM system,” in Proc. IEEE
GLOBECOM, 2005, pp. 1423-1428.
[13] J. A. Prakash and G. R. Reddy, “Optimized generalized gaussian
pulse (OGGP) based prototype filter design for filter bank multi
carrier (FBMC) systems,” in Proc. First International Conference
on Computational Systems and Communications, Trivandrum,
India, 2014, pp. 317-322.
[15] C. H. Wilcox, “The synthesis problem for radar ambiguity
functions,” in Radar and Sonar I, New York: Springer, 1991, pp.
252-253.
[16] P. Amini, R. R. Chen, and B. Farhang-Boroujeny, “Filterbank
multicarrier communications for underwater acoustic channels,”
IEEE Journal of Oceanic Engineering, pp. 1-16, 2014.
[17] W. K. N. Derrick, S. L. Ernest, and S. Robert, “Energy efficient
resource allocation in OFDMA systems with hybrid energy
harvesting base station,” IEEE Trans. on Wireless Communications,
vol. 12, no. 7, pp. 3412-3427, 2013.
[18] Z. Q. Luo, W. K. Ma, S. A. Man-Cho, Y. Y. Ye, and S. Z. Zhang,
“Semidefinite relaxation of quadratic optimization problems,”
IEEE Signal Processing Magazine, vol. 27, no. 3, pp. 20-34, 2010.
[19] M. Grant, S. Boyd, and Y. Ye. (2009). CVX: MATLAB software
for disciplined convex programming. [Online]. Available:
http://www.stanford.edu/~boyd/cvx
Journal of Communications Vol. 10, No. 5, May 2015
306©2015 Journal of Communications
[14] D. Chen, D. M. Qu, T. Jiang, and Y. J. He. “Prototype filter
optimization to minimize stopband energy with NPR constraint for
filter bank multicarrier modulation systems,” IEEE Trans. on
Signal Processing, vol. 61, no. 1, pp. 159-170, 2013.
Yu Zhao received the B.S. and M.S. degrees
in 2009 and 2011, respectively, from Air
Force Engineering University, Xi’an. He is
currently working toward the Ph.D. degree in
the Air and Missile Defense College. His
research interests include information theory
and multicarrier modulation techniques.
Xihong Chen received the M.S. degree in
communication engineering from Xidian
University, Xi’an, in 1992 and the Ph.D.
degree from Missile College of Air Force
Engineering University in 2010. He is
currently a professor with Air and Missile
Defense College, AFEU, Xi’an. His research
interests include information theory,
information security and signal processing.
Qun Zhang
received the M.S. degree in
mathematics from Shanxi Normal University,
Xi’an, in 1988, and the Ph.D. degree in
electrical engineering from Xidian University,
Xi’an,
in 2001.
He is currently a professor
with Information and Navigation College,
AFEU, Xi’an. His
research interests include
signal processing, clutter
suppression, and its
application in SAR and ISAR.
Journal of Communications Vol. 10, No. 5, May 2015
307©2015 Journal of Communications