chapter 6: correlation functions
TRANSCRIPT
Notes and figures are based on or taken from materials in the course textbook: Probabilistic Methods of Signal and System Analysis (3rd ed.) by George R. Cooper and Clare D. McGillem; Oxford Press, 1999. ISBN: 0-19-512354-9.
B.J. Bazuin, Spring 2015 1 of 27 ECE 3800
Chapter 6: Correlation Functions
6-1 Introduction
6-2 Example: Autocorrelation Function of a Binary Process
6-3 Properties of Autocorrelation Functions
6-4 Measurement of Autocorrelation Functions
6-5 Examples of Autocorrelation Functions
6-6 Crosscorrelation Functions
6-7 Properties of Crosscorrelation Functions
6-8 Example and Applications of Crosscorrelation Functions
6-9 Correlation Matrices for Sampled Functions
Concepts
Autocorrelation o of a Binary/Bipolar Process o of a Sinusoidal Process
Linear Transformations Properties of Autocorrelation Functions Measurement of Autocorrelation Functions An Anthology of Signals Crosscorrelation Properties of Crosscorrelation Examples and Applications of Crosscorrelation Correlation Matrices For Sampled Functions
Notes and figures are based on or taken from materials in the course textbook: Probabilistic Methods of Signal and System Analysis (3rd ed.) by George R. Cooper and Clare D. McGillem; Oxford Press, 1999. ISBN: 0-19-512354-9.
B.J. Bazuin, Spring 2015 2 of 27 ECE 3800
5-5.3 A random process has sample functions of the form
n
tnTtfAtX 0
where A and T are constants and t0 is a random variable that is uniformly distributed between 0 and T. The function f(t) is defined by
201 Tttf
and zero elsewhere.
Also Note: TtT
tpdf 00 01
a. Find the mean and second moment.
00 tnTtfEAEtnTtfAEtXEn
0tnTtfEAtXE
For any time t, there is a probability that f(t) is present or absent, based on the random variable t0. As t0 is uniformly distributed across the bit period T and f(t) is present for exactly ½ of the period,
2
1
2
110
0
00 T
Tdt
TtftnTtfE
T
and
2
1 AtXE
nn
tnTtfAEtnTtfAEtXE 02
2
02
022 tnTtfEAtXE
As previously stated
2
122 AtXE
Notes and figures are based on or taken from materials in the course textbook: Probabilistic Methods of Signal and System Analysis (3rd ed.) by George R. Cooper and Clare D. McGillem; Oxford Press, 1999. ISBN: 0-19-512354-9.
B.J. Bazuin, Spring 2015 3 of 27 ECE 3800
b. Find the time average and squared average.
0
0
0
1tT
t n
dttnTtfAT
tX
22
11 2
00
AT
T
Adt
T
AdttfA
TtX
TT
and
0
0
2
02 1
tT
t n
dttnTtfAT
tX
22
11 222
0
2
0
22 AT
T
Adt
T
AdttfA
TtX
TT
c. Can this process be stationary?
Yes, the mean and variance are not functions of time. Therefore, we expect it to be wide-sense stationary.
d. Can this process be ergodic?
Based on the information provide, there is no specific reason to suggest that the process is not ergodic.
Notes and figures are based on or taken from materials in the course textbook: Probabilistic Methods of Signal and System Analysis (3rd ed.) by George R. Cooper and Clare D. McGillem; Oxford Press, 1999. ISBN: 0-19-512354-9.
B.J. Bazuin, Spring 2015 4 of 27 ECE 3800
Chapter 6: Correlation Functions
The Autocorrelation Function
For a sample function defined by samples in time of a random process, how alike are the different samples?
Define: 11 tXX and 22 tXX
The autocorrelation is defined as:
2121, tXtXEttRXX
The above function is valid for all processes, stationary and non-stationary.
For WSS processes:
XXXX RtXtXEttR 21,
If the process is ergodic, the time average is equivalent to the probabilistic expectation, or
txtxdttxtx
T
T
TT
XX 2
1lim
and
XXXX R
Notes and figures are based on or taken from materials in the course textbook: Probabilistic Methods of Signal and System Analysis (3rd ed.) by George R. Cooper and Clare D. McGillem; Oxford Press, 1999. ISBN: 0-19-512354-9.
B.J. Bazuin, Spring 2015 5 of 27 ECE 3800
Example: tfAtx 2sin for A a uniformly distributed random variable 2,2A
212121 2sin2sin, tfAtfAEtXtXEttRXX
2121
22121 2cos2cos
2
1, ttfttfAEtXtXEttRXX
21212
21 2cos2cos2
1, ttfttfAEttRXX
for 12 tt
21
2
21 2cos2cos122
1, ttffttRXX
2121 2cos2cos24
16, ttffttRXX
A non-stationary process
The time based formulation:
txtxdttxtx
T
T
TT
XX 2
1lim
T
TT
XX dttfAtfAT
2sin2sin2
1lim
T
TT
XX dttffT
A 22cos2cos2
1
2
1lim2
T
TT
XX dttfT
Af
A 22cos2
1lim
22cos
2
22
fA
fA
XX 2cos2
2cos2
22
Acceptable, but the R.V. is still present?!
ff
AEE XX 2cos
24
162cos
2
2
Notes and figures are based on or taken from materials in the course textbook: Probabilistic Methods of Signal and System Analysis (3rd ed.) by George R. Cooper and Clare D. McGillem; Oxford Press, 1999. ISBN: 0-19-512354-9.
B.J. Bazuin, Spring 2015 6 of 27 ECE 3800
Example: tfAtx 2sin for a uniformly distributed random variable 2,0
212121 2sin2sin, tfAtfAEtXtXEttRXX
22cos2cos2
1, 2121
22121 ttfttfEAtXtXEttRXX
22cos2
2cos2
, 21
2
21
2
2121 ttfEA
ttfA
tXtXEttRXX
Of note is that the phase need only be uniformly distributed over 0 to π in the previous step!
21
2
2121 2cos2
, ttfA
tXtXEttRXX
for 12 tt
fA
RXX 2cos2
2
but
fA
RR XXXX 2cos2
2
Assuming a uniformly distributed random phase “simplifies the problem” …
Also of note, if the amplitude is an independent random variable, then
fAE
RXX 2cos2
2
The time based formulation:
txtxdttxtx
T
T
TT
XX 2
1lim
T
TT
XX dttfAtfAT
2sin2sin2
1lim
T
TT
XX dttffT
A 222cos2cos2
1lim
2
2
fA
XX 2cos2
2
Notes and figures are based on or taken from materials in the course textbook: Probabilistic Methods of Signal and System Analysis (3rd ed.) by George R. Cooper and Clare D. McGillem; Oxford Press, 1999. ISBN: 0-19-512354-9.
B.J. Bazuin, Spring 2015 7 of 27 ECE 3800
Example:
T
ttrectBtx 0 for B =+/-A with probability p and (1-p) and t0 a uniformly
distributed random variable
2,
20
TTt . Assume B and t0 are independent.
T
ttrectB
T
ttrectBEtXtXEttRXX
01012121 ,
T
ttrect
T
ttrectBEtXtXEttRXX
010122121 ,
As the RV are independent
T
ttrect
T
ttrectEBEtXtXEttRXX
020122121 ,
T
ttrect
T
ttrectEpApAttRXX
02012221 1,
2
2
002012
21
1,
T
TXX dt
TT
ttrect
T
ttrectAttR
For 21 0 tandt
2
2
002 1
1,0
T
TXX dt
TT
trectAR
The integral can be recognized as being a triangle, extending from –T to T and zero everywhere else.
T
triARXX
2
T
TTT
A
TTT
A
T
RXX
,0
0,1
0,1
,0
2
2
Notes and figures are based on or taken from materials in the course textbook: Probabilistic Methods of Signal and System Analysis (3rd ed.) by George R. Cooper and Clare D. McGillem; Oxford Press, 1999. ISBN: 0-19-512354-9.
B.J. Bazuin, Spring 2015 8 of 27 ECE 3800
The time based formulation:
txtxdttxtx
T
T
TT
XX 2
1lim
C
CC
XX dtT
ttrectB
T
ttrectB
C00
2
1lim
A change in variable for the integral 0ttt . And only integrate over the finite interval T.
2
2
2 11
T
TXX dt
T
trect
TB
T
ttriBXX
2
T
ttriBEE XX
2
T
ttripApAE XX
122
T
ttriAE XX
2
Notes and figures are based on or taken from materials in the course textbook: Probabilistic Methods of Signal and System Analysis (3rd ed.) by George R. Cooper and Clare D. McGillem; Oxford Press, 1999. ISBN: 0-19-512354-9.
B.J. Bazuin, Spring 2015 9 of 27 ECE 3800
Properties of Autocorrelation Functions
1) 220 XXERXX or 20 txXX
the mean squared value of the random process can be obtained by observing the zeroth lag of the autocorrelation function.
2) XXXX RR
The autocorrelation function is an even function in time. Only positive (or negative) needs to be computed for an ergodic, WSS random process.
3) 0XXXX RR
The autocorrelation function is a maximum at 0. For periodic functions, other values may equal the zeroth lag, but never be larger.
4) If X has a DC component, then Rxx has a constant factor.
tNXtX
NNXX RXR 2
Note that the mean value can be computed from the autocorrelation function constants!
5) If X has a periodic component, then Rxx will also have a periodic component of the same period.
Think of:
20,cos twAtX
where A and w are known constants and theta is a uniform random variable.
wA
tXtXERXX cos2
2
Notes and figures are based on or taken from materials in the course textbook: Probabilistic Methods of Signal and System Analysis (3rd ed.) by George R. Cooper and Clare D. McGillem; Oxford Press, 1999. ISBN: 0-19-512354-9.
B.J. Bazuin, Spring 2015 10 of 27 ECE 3800
5b) For signals that are the sum of independent random variable, the autocorrelation is the sum of the individual autocorrelation functions.
tYtXtW
YXYYXXWW RRR 2
For non-zero mean functions, (let w, x, y be zero mean and W, X, Y have a mean)
YXYYXXWW RRR 2
YXYyyXxxWwwWW RRRR 2222
222 2 YYXXyyxxWwwWW RRRR
22YXyyxxWwwWW RRRR
Then we have
22YXW
yyxxww RRR
6) If X is ergodic and zero mean and has no periodic component, then
0lim
XXR
No good proof provided.
The logical argument is that eventually new samples of a non-deterministic random process will become uncorrelated. Once the samples are uncorrelated, the autocorrelation goes to zero.
7) Autocorrelation functions can not have an arbitrary shape. One way of specifying shapes permissible is in terms of the Fourier transform of the autocorrelation function. That is, if
dtjwtRR XXXX exp
then the restriction states that
wallforRXX 0
Notes and figures are based on or taken from materials in the course textbook: Probabilistic Methods of Signal and System Analysis (3rd ed.) by George R. Cooper and Clare D. McGillem; Oxford Press, 1999. ISBN: 0-19-512354-9.
B.J. Bazuin, Spring 2015 11 of 27 ECE 3800
Additional concepts that are useful … dealing with constant:
tNatX
tNatNaEtXtXERXX
NNXX RatNtNEaR 22
tYatX
tYatYaERXX
tYtYtYatYaaERXX2
tYtYEtYEatYEaaRXX2
YYXX RtYEaaR 22
Exercise 6-1.1
A random process has a sample function of the form
else
tAtX
,0
10,
where A is a random variable that is uniformly distributed from 0 to 10.
Find the autocorrelation of the process.
100,10
1 aforaf
Using
2121, tXtXEttRXX
1,0,, 212
21 ttforAEttRXX
1,0,10
1, 21
10
0
221 ttfordaattRXX
1,0,30
, 21
10
0
3
21 ttfora
ttRXX
1,0,3
100
30
1000, 2121 ttforttRXX
Notes and figures are based on or taken from materials in the course textbook: Probabilistic Methods of Signal and System Analysis (3rd ed.) by George R. Cooper and Clare D. McGillem; Oxford Press, 1999. ISBN: 0-19-512354-9.
B.J. Bazuin, Spring 2015 12 of 27 ECE 3800
Exercise 6-1.2
Define a random variable Z(t) as
1 tXtXtZ
where X(t) is a sample function from a stationary random process whose autocorrelation function is
2exp XXR
Write an expressions for the autocorrelation function of Z(t).
2121, tZtZEttRZZ
12211121, tXtXtXtXEttRZZ
12112111212121, tXtXtXtXtXtXtXtXEttRZZ
1211211
1212121 ,
tXtXEtXtXE
tXtXEtXtXEttRZZ
Note: based on the autocorrelation definition, 12 tt
121121121221, ttRttRttRttRttR XXXXXXXXZZ
1121 2, XXXXXXZZ RRRttR
21
21
221 expexpexp2, ttRZZ
Note that for this example, the autocorrelation is strictly a function of tau and not t1 or t2. As a result, we expect that the random variable is stationary.
Helpful in defining a stationary random process:
A stationary random variable will have an autocorrelation that is dependent on the time difference, not the absolute times.
Notes and figures are based on or taken from materials in the course textbook: Probabilistic Methods of Signal and System Analysis (3rd ed.) by George R. Cooper and Clare D. McGillem; Oxford Press, 1999. ISBN: 0-19-512354-9.
B.J. Bazuin, Spring 2015 13 of 27 ECE 3800
Autocorrelation of a Binary Process
Binary communications signals are discrete, non-deterministic signals with plenty of random variables involved in their generations.
Nominally, the binary bit last for a defined period of time, ta, but initially occurs at a random delay (uniformly distributed across one bit interval 0<=t0<ta). The amplitude may be a fixed constant or a random variable.
ta
t0A
-A
bit(0) bit(1) bit(2) bit(4)bit(3)
The bit values are independent, identically distributed with probability p that amplitude is A and q=1-p that amplitude is –A.
aa
ttfort
tpdf 00 0,1
Determine the autocorrelation of the bipolar binary sequence, assuming p=0.5.
k a
aa
k t
tktttrectatx
20
2121, tXtXEttRXX
For samples more than one period apart, attt 21 , we must consider
ApApApApApApApApaaE jk 1111
222 112 ppppAaaE jk
144 22 ppAaaE jk
For p=0.5 0144 22 ppAaaE jk
Notes and figures are based on or taken from materials in the course textbook: Probabilistic Methods of Signal and System Analysis (3rd ed.) by George R. Cooper and Clare D. McGillem; Oxford Press, 1999. ISBN: 0-19-512354-9.
B.J. Bazuin, Spring 2015 14 of 27 ECE 3800
For samples within one period, attt 21 ,
222 1 AApApaaE kk
0144 221 ppAaaE kk
there are two regions to consider, the sample bit overlapping and the area of the next bit. But the overlapping area … should be triangular. Therefore
0,11 2
2
2
2
1
a
t
tkk
a
t
tkk
aXX tfordtaaE
tdtaaE
tR
a
a
a
a
a
t
tkk
a
t
tkk
aXX tfordtaaE
tdtaaE
tR
a
a
a
a
0,11 2
2
1
2
2
or
0,11 2
2
2
a
t
taXX tfordt
tAR
a
a
a
t
taXX tfordt
tAR
a
a
0,11 2
2
2
Therefore
aa
a
aa
a
XX
tfort
tA
tfort
tA
R
0,
0,
2
2
or recognizing the structure
aaa
XX ttfort
AR
,12
This is simply a triangular function with maximum of A2, extending for a full bit period in both time directions.
Notes and figures are based on or taken from materials in the course textbook: Probabilistic Methods of Signal and System Analysis (3rd ed.) by George R. Cooper and Clare D. McGillem; Oxford Press, 1999. ISBN: 0-19-512354-9.
B.J. Bazuin, Spring 2015 15 of 27 ECE 3800
For unequal bit probability
a
aaaa
a
XX
tforppA
ttfort
ppt
tA
R
,144
,144
22
22
As there are more of one bit or the other, there is always a positive correlation between bits (the curve is a minimum for p=0.5), that peaks to A2 at = 0.
Note that if the amplitude is a random variable, the expected value of the bits must be further evaluated. Such as,
22 kk aaE
21 kk aaE
In general, the autocorrelation of communications signal waveforms is important, particularly when we discuss the power spectral density later in the textbook.
Notes and figures are based on or taken from materials in the course textbook: Probabilistic Methods of Signal and System Analysis (3rd ed.) by George R. Cooper and Clare D. McGillem; Oxford Press, 1999. ISBN: 0-19-512354-9.
B.J. Bazuin, Spring 2015 16 of 27 ECE 3800
Examples of discrete waveforms used for communications, signal processing, controls, etc.
(a) Unipolar RZ & NRZ, (b) Polar RZ & NRZ , (c) Bipolar NRZ , (d) Split-phase Manchester, (e) Polar quaternary NRZ.
From Cahp. 11: A. Bruce Carlson, P.B. Crilly, Communication Systems, 5th ed., McGraw-Hill, 2010. ISBN: 978-0-07-338040-7
In general, a periodic bipolar “pulse” that is shorter in duration than the pulse period will have the autocorrelation function
wwwp
wXX ttfor
tt
tAR
,12
for a tw width pulse existing in a tp time period, assuming that positive and negative levels are equally likely.
Notes and figures are based on or taken from materials in the course textbook: Probabilistic Methods of Signal and System Analysis (3rd ed.) by George R. Cooper and Clare D. McGillem; Oxford Press, 1999. ISBN: 0-19-512354-9.
B.J. Bazuin, Spring 2015 17 of 27 ECE 3800
Exercise 6-3.1
a) An ergodic random process has an autocorrelation function of the form
1610cos164exp9 XXR
Find the mean-square value, the mean value, and the variance of the process.
The mean-square (2nd moment) is
222 41161690 XXRXE
The constant portion of the autocorrelation represents the square of the mean. Therefore
1622 XE and 4
Finally, the variance can be computed as,
2516410 2222 XXRXEXE
b) An ergodic random process has an autocorrelation function of the form
1
642
2
XXR
Find the mean-square value, the mean value, and the variance of the process.
The mean-square (2nd moment) is
222 61
60 XXRXE
The constant portion of the autocorrelation represents the square of the mean. Therefore
41
4
1
64lim
2
222
t
XE and 2
Finally, the variance can be computed as,
2460 2222 XXRXEXE
Notes and figures are based on or taken from materials in the course textbook: Probabilistic Methods of Signal and System Analysis (3rd ed.) by George R. Cooper and Clare D. McGillem; Oxford Press, 1999. ISBN: 0-19-512354-9.
B.J. Bazuin, Spring 2015 18 of 27 ECE 3800
The Crosscorrelation Function
For a two sample function defined by samples in time of two random processes, how alike are the different samples?
Define: 11 tXX and 22 tYY
The cross-correlation is defined as:
2121212121 ,, yxfyxdydxYXEttRXY
2121212121 ,, xyfxydxdyXYEttRYX
The above function is valid for all processes, jointly stationary and non-stationary.
For jointly WSS processes:
XYXY RtYtXEttR 21,
YXYX RtXtYEttR 21,
Note: the order of the subscripts is important for cross-correlation!
If the processes are jointly ergodic, the time average is equivalent to the probabilistic expectation, or
tytxdttytx
T
T
TT
XY 2
1lim
txtydttxty
T
T
TT
YX 2
1lim
and
XYXY R
YXYX R
Notes and figures are based on or taken from materials in the course textbook: Probabilistic Methods of Signal and System Analysis (3rd ed.) by George R. Cooper and Clare D. McGillem; Oxford Press, 1999. ISBN: 0-19-512354-9.
B.J. Bazuin, Spring 2015 19 of 27 ECE 3800
Properties of Crosscorrelation Functions
1) The properties of the zoreth lag have no particular significance and do not represent mean-square values. It is true that the “ordered” crosscorrelations must be equal at 0. .
00 YXXY RR or 00 YXXY
2) Crosscorrelation functions are not generally even functions. However, there is an antisymmetry to the ordered crosscorrelations:
YXXY RR
For
tytxdttytx
T
T
TT
XY 2
1lim
Substitute t
yxdyxT
T
TT
XY
2
1lim
YX
T
TT
XY xydxyT2
1lim
3) The crosscorrelation does not necessarily have its maximum at the zeroth lag. This makes sense if you are correlating a signal with a timed delayed version of itself. The crosscorrelation should be a maximum when the lag equals the time delay!
It can be shown however that 00 XXXXXY RRR
As a note, the crosscorrelation may not achieve the maximum anywhere …
4) If X and Y are statistically independent, then the ordering is not important
YXtYEtXEtYtXERXY
and
YXXY RYXR
Notes and figures are based on or taken from materials in the course textbook: Probabilistic Methods of Signal and System Analysis (3rd ed.) by George R. Cooper and Clare D. McGillem; Oxford Press, 1999. ISBN: 0-19-512354-9.
B.J. Bazuin, Spring 2015 20 of 27 ECE 3800
5) If X is a stationary random process and is differentiable with respect to time, the crosscorrelation of the signal and it’s derivative is given by
d
dRR XX
XX
Defining derivation as a limit:
e
tXetXX
e
0lim
and the crosscorrelation
e
tXetXtXEtXtXER
eXX
0
lim
e
tXtXetXtXER
eXX
0lim
e
tXtXEetXtXER
eXX
0lim
e
ReRR XXXX
eXX
0lim
d
dRR XX
XX
Similarly,
2
2
d
RdR XX
XX
That is all the properties for the crosscorrelation function.
Notes and figures are based on or taken from materials in the course textbook: Probabilistic Methods of Signal and System Analysis (3rd ed.) by George R. Cooper and Clare D. McGillem; Oxford Press, 1999. ISBN: 0-19-512354-9.
B.J. Bazuin, Spring 2015 21 of 27 ECE 3800
Example: A delayed signal in noise … cross correlated with the original signal …
tX
tNttXatY d
Find
tYtXERXY
Then
tNttXatXER dXY
XNdXXXY RtRaR
For noise independent of the signal,
NXXN RNXR
and therefore
NXtRaR dXXXY
For either X or N a zero mean process, we would have
dXXXY tRaR
Does this suggest a way to determine the time delay based on what you know of the autocorrelation?
The maximum of the autocorrelation is at zero; therefore, the maximum of the crosscorrelation can reveal the time delay!
any electronic system based on the time delay of a known, transmitted signal pulse …
Notes and figures are based on or taken from materials in the course textbook: Probabilistic Methods of Signal and System Analysis (3rd ed.) by George R. Cooper and Clare D. McGillem; Oxford Press, 1999. ISBN: 0-19-512354-9.
B.J. Bazuin, Spring 2015 22 of 27 ECE 3800
Back to Autocorrelation examples, using time a time average Example: Rect. Function
From Dr. Severance’s notes;
k
k T
kTttctAtX 0Re
where A(k) are zero mean, identically distributed, independent R.V., t0 is a R.V. uniformly distributed over possible bit period T, and T is a constant. k describes the bit positions.
j
j
k
kXX T
jTttctA
T
kTttctAEttR 00 ReRe,
T
jTttct
T
kTttctAAEttR jk
jk
XX00 ReRe,
T
jTttct
T
kTttctEAAEttR jk
jk
XX00 ReRe,
but
jkfor
jkforAEAAE jk
0
2
Therefore (one summation goes away)
T
kTttct
T
kTttctEAEttR
k
XX00 ReRe, 2
Now using the time average concept
txtxdttxtx
T
T
TT
XX
2
2
1lim
Establishing the time average as the average of the kth bit, (notice that the delay is incorporated into the integral, and assuming that the signal is Wide-Sense Stationary)
0
0
2
2
2
Re11
tT
tT
XX dtT
tct
TAE
Notes and figures are based on or taken from materials in the course textbook: Probabilistic Methods of Signal and System Analysis (3rd ed.) by George R. Cooper and Clare D. McGillem; Oxford Press, 1999. ISBN: 0-19-512354-9.
B.J. Bazuin, Spring 2015 23 of 27 ECE 3800
for T>>=0
0
0
2
0
0
2 2
2
2
2
111
1 tT
tT
tT
tT
XX tT
AEdtT
AE
TT
AEtTtTT
AEXX1
221
2200
0,12
for
TAEXX
Due to symmetry, XXXX , we have
TAEXX
12
Notes and figures are based on or taken from materials in the course textbook: Probabilistic Methods of Signal and System Analysis (3rd ed.) by George R. Cooper and Clare D. McGillem; Oxford Press, 1999. ISBN: 0-19-512354-9.
B.J. Bazuin, Spring 2015 24 of 27 ECE 3800
Example: Rect. Function that does not extend for the full period
k
k T
kTttctAtX
0Re
where A(k) are zero mean, identically distributed, independent R.V., t0 is a R.V. uniformly distributed over possible bit period T, and T and alpha are constants. k describes the bit positions.
From the previous example, the average value during the time period was defined as:
0
0
2
2
2
Re11
tT
tT
XX dtT
tct
TAE
For this example, the period remains the same, but the integration period due to the rect functions change to:
0
0
2
2
2
Re11
tT
tT
XX dtT
tct
TAE
for T>>=0
0
0
2
0
0
2 2
2
2
2
111
1 tT
tT
tT
tT
XX tT
AEdtT
AE
TT
AEtTtTT
AEXX1
221
2200
0,11 22
for
TAE
TT
TAEXX
Due to symmetry, we have
TAEXX
12
Notes and figures are based on or taken from materials in the course textbook: Probabilistic Methods of Signal and System Analysis (3rd ed.) by George R. Cooper and Clare D. McGillem; Oxford Press, 1999. ISBN: 0-19-512354-9.
B.J. Bazuin, Spring 2015 25 of 27 ECE 3800
Measurement of the Autocorrelation Function
We love to use time average for everything. For wide-sense stationary, ergodic random processes, time average are equivalent to statistical or probability based values.
txtxdttxtx
T
T
TT
XX 2
1lim
Using this fact, how can we use short-term time averages to generate auto- or cross-correlation functions?
An estimate of the autocorrelation is defined as:
T
XX dttxtxT
R
0
1ˆ
Note that the time average is performed across as much of the signal that is available after the time shift by tau.
In most practical cases, function is performed in terms of digital samples taken at specific time intervals,
t . For tau based on the available time step, k, with N equating to the available time interval, we have:
kN
i
XX ttktixtixtktN
tkR
01
1ˆ
kN
i
XXXX kixixkN
kRtkR
01
1ˆˆ
In computing this autocorrelation, the initial weighting term approaches 1 when k=N. At this point the entire summation consists of one point and is therefore a poor estimate of the autocorrelation. For useful results, k<<N!
Notes and figures are based on or taken from materials in the course textbook: Probabilistic Methods of Signal and System Analysis (3rd ed.) by George R. Cooper and Clare D. McGillem; Oxford Press, 1999. ISBN: 0-19-512354-9.
B.J. Bazuin, Spring 2015 26 of 27 ECE 3800
It can be shown that the estimated autocorrelation is equivalent to the actual autocorrelation; therefore, this is an unbiased estimate.
kN
i
XXXX kixixkN
EkREtkRE
01
1ˆˆ
tkRkNkN
tkRkN
kixixEkN
tkRE
XX
kN
i
XX
kN
i
XX
11
1
1
1
1
1ˆ
0
0
tkRtkRE XXXX ˆ
As noted, the validity of each of the summed autocorrelation lags can and should be brought into question as k approaches N. As a result, a biased estimate of the autocorrelation is commonly used. The biased estimate is defined as:
kN
i
XX kixixN
kR
01
1~
Here, a constant weight instead of one based on the number of elements summed is used. This estimate has the property that the estimated autocorrelation should decrease as k approaches N.
The expected value of this estimate can be shown to be
tkRN
nkRE XXXX
11
~
The variance of this estimate can be shown to be (math not done at this level)
M
Mk
XXXX tkRN
kRVar 22~
This equation can be used to estimate the number of time samples needed for a useful estimate.
Notes and figures are based on or taken from materials in the course textbook: Probabilistic Methods of Signal and System Analysis (3rd ed.) by George R. Cooper and Clare D. McGillem; Oxford Press, 1999. ISBN: 0-19-512354-9.
B.J. Bazuin, Spring 2015 27 of 27 ECE 3800
Matlab Example %% % Figure 6_2 % clear; close all; nsamples = 1000; n=(0:1:nsamples-1)'; %% % rect rect_length = 100; x = zeros(nsamples,1); x(100+(1:rect_length))=1; % Scale to make the maximum unity Rxx1 = xcorr(x)/rect_length; Rxx2 = xcorr(x,'unbiased'); nn = -(nsamples-1):1:(nsamples-1); figure plot(n,x) figure plot(nn,Rxx1,nn,Rxx2) legend('Raw','Unbiased')
-1000 -800 -600 -400 -200 0 200 400 600 800 1000-0.2
0
0.2
0.4
0.6
0.8
1
1.2
Raw
Unbiased