Classification of LTI Discrete-

Discrete- Classification of LTI Discrete-

Discrete-
Time Systems Time Systems
Based on the Output Calculation Process Based on the Coefficients -
• Nonrecursive System - Here the output can • Real Discrete-Time System - The impulse
be calculated sequentially, knowing only response samples are real valued
the present and past input samples • Complex Discrete-Time System - The
• Recursive System - Here the output impulse response samples are complex
computation involves past output samples in valued
addition to the present and past input
samples
25 Copyright © 2005, S. K. Mitra 26 Copyright © 2005, S. K. Mitra

Correlation of Signals Correlation of Signals

• There are applications where it is necessary • For example, in digital communications, a
to compare one reference signal with one or set of data symbols are represented by a set
more signals to determine the similarity of unique discrete-time sequences
between the pair and to determine additional • If one of these sequences has been
information based on the similarity transmitted, the receiver has to determine
which particular sequence has been received
by comparing the received signal with every
member of possible sequences from the set
27 Copyright © 2005, S. K. Mitra 28 Copyright © 2005, S. K. Mitra

Correlation of Signals Correlation of Signals

• Similarly, in radar and sonar applications,
the received signal reflected from the target Definitions
is a delayed version of the transmitted • A measure of similarity between a pair of
signal and by measuring the delay, one can energy signals, x[n] and y[n], is given by the
determine the location of the target cross-correlation sequence rxy [�] defined by
• The detection problem gets more �
complicated in practice, as often the rxy [�] � � x[n] y[n � �], � � 0, � 1, � 2, ...
n � ��
received signal is corrupted by additive
• The parameter � called lag, indicates the
random noise
time-shift between the pair of signals
29 Copyright © 2005, S. K. Mitra 30 Copyright © 2005, S. K. Mitra

5
 Correlation of Signals Correlation of Signals
• y[n] is said to be shifted by � samples to the • If y[n] is made the reference signal and shift
right with respect to the reference sequence x[n] with respect to y[n], then the
x[n] for positive values of �, and shifted by � corresponding cross-correlation sequence is
samples to the left for negative values of given by
• The ordering of the subscripts xy in the ryx [�] � ��
n � �� y[n]x[n � � ]
definition of rxy [�] specifies that x[n] is the
reference sequence which remains fixed in � ��
m � �� y[m � �]x[m] � rxy [�� ]
time while y[n] is being shifted with respect
to x[n] • Thus, ryx [�] is obtained by time-reversing
rxy [�]
31 Copyright © 2005, S. K. Mitra 32 Copyright © 2005, S. K. Mitra

Correlation of Signals Correlation of Signals

• The autocorrelation sequence of x[n] is • From the relation ryx [�] � rxy [��] it follows
given by that rxx [�] � rxx [��] implying that rxx [�] is
rxx [�] � ��
n � �� x[n]x[n � �]
an even function for real x[n]
obtained by setting y[n] = x[n] in the • An examination of
definition of the cross-correlation sequence rxy [�] � ��n � �� x[n] y[n � �]
rxy [�] reveals that the expression for the cross-
• Note: rxx [0] � �� 2
n � �� x [n] � E x , the energy
correlation looks quite similar to that of the
of the signal x[n] linear convolution
33 Copyright © 2005, S. K. Mitra 34 Copyright © 2005, S. K. Mitra

Correlation of Signals Correlation of Signals

• This similarity is much clearer if we rewrite
the expression for the cross-correlation as • Likewise, the autocorrelation of x[n] can be
rxy [�] � ��
n � �� x[n] y[�(� � n)] � x[� ] * y[��]
computed by processing x[n] with an LTI
discrete-time system of impulse response
• The cross-correlation of y[n] with the x[�n]
reference signal x[n] can be computed by x[n ] x[ � n ] rxx [ n ]
processing x[n] with an LTI discrete-time
system of impulse response y[� n]
x[n ] y[ � n ] rxy [n ]

35 Copyright © 2005, S. K. Mitra 36 Copyright © 2005, S. K. Mitra

6
 Properties of Autocorrelation and Properties of Autocorrelation and
Cross-
Cross-correlation Sequences Cross-
Cross-correlation Sequences
• Consider two finite-energy sequences x[n] • Thus
and y[n] a 2rxx [0] � 2a rxy [�] � ryy [0] � 0
• The energy of the combined sequence where rxx [0] � Ex � 0 and ryy [0] � E y � 0
a x[n] � y[n � �] is also finite and
• We can rewrite the equation on the previous
nonnegative, i.e., slide as
��n��� (a x[n] � y[n � �])2 � a 2 ��n��� x 2 [n]
�a 1��� rxx [�] rxy [0]�� ��1a �� � 0
r [0] r [�]
� 2a �� � 2
n � �� x[n] y[n � �] � �n � �� y [n � � ] � 0
� xy yy �
for any finite value of a
37 Copyright © 2005, S. K. Mitra 38 Copyright © 2005, S. K. Mitra

Properties of Autocorrelation and Properties of Autocorrelation and

Cross-
Cross-correlation Sequences Cross-
Cross-correlation Sequences
• Or, in other words, the matrix • The last inequality on the previous slide
�rxx [0] rxy [�]� provides an upper bound for the cross-
� r [�] r [0]� correlation samples
� xy yy �
is positive semidefinite • If we set y[n] = x[n], then the inequality
2 reduces to
• rxx [0]ryy [0] � rxy [� ] � 0 | rxy [�] | � rxx [0] � Ex
or, equivalently,
| rxy [�] | � rxx [0]ryy [0] � E xE y
39 Copyright © 2005, S. K. Mitra 40 Copyright © 2005, S. K. Mitra

Properties of Autocorrelation and Properties of Autocorrelation and

Cross-
Cross-correlation Sequences Cross-
Cross-correlation Sequences
• Thus, at zero lag ( � � 0), the sample value • Therefore
of the autocorrelation sequence has its ExE y � b 2Ex2 � bEx
maximum value • Using the above result in
• Now consider the case | rxy [�] | � rxx [0]ryy [0] � E xE y
y[n] � �b x[n � N ]
we get
where N is an integer and b > 0 is an � b rxx [0] � rxy [�] � b rxx [0]
arbitrary number
• In this case E y � b 2E x
41 Copyright © 2005, S. K. Mitra 42 Copyright © 2005, S. K. Mitra

7
 Correlation Computation Correlation Computation
Using MATLAB Using MATLAB
• The cross-correlation and autocorrelation • The cross-correlation sequence rxy [n]
sequences can easily be computed using computed using Program 2_7 of text is
MATLAB plotted below
• Example - Consider the two finite-length 30

sequences 20

x[n] � �1 3 � 2 1 2 � 1 4 4 2�

Amplitude
10

y[n] � �2 � 1 4 1 � 2 3� -10
-4 -2 0 2 4 6 8
Lag index

43 Copyright © 2005, S. K. Mitra 44 Copyright © 2005, S. K. Mitra

Correlation Computation Correlation Computation

Using MATLAB Using MATLAB
• The autocorrelation sequence rxx [�] • The plot below shows the cross-correlation
computed using Program 2_7 is shown below of x[n] and y[n] � x[n � N ] for N = 4
• Note: At zero lag, rxx [0] is the maximum • Note: The peak of the cross-correlation is
60
precisely the value of the delay N
60
40
40
Amplitude

Amplitude

20
20
0
0
-20
-5 0 5 -20
Lag index -10 -5 0 5
45 Copyright © 2005, S. K. Mitra 46 Lag index
Copyright © 2005, S. K. Mitra

Correlation Computation
Correlation Computation
Using MATLAB
Using MATLAB
• The plot below shows the autocorrelation of
x[n] corrupted with an additive random • The autocorrelation and the cross-
noise generated using the function randn correlation can also be computed using the
function xcorr
• Note: The autocorrelation still exhibits a
peak at zero lag • However, the correlation sequences
80 generated using this function are the time-
60
reversed version of those generated using
Amplitude

40
Programs 2_7 and 2_8
20

0
-5 0 5
47 Lag index Copyright © 2005, S. K. Mitra 48 Copyright © 2005, S. K. Mitra

8
 Normalized Forms of Correlation Computation for
Correlation
Power Signals
• Normalized forms of autocorrelation and
cross-correlation are given by • The cross-correlation sequence for a pair of
r [�] rxy [�] power signals, x[n] and y[n], is defined as
� xx [�] � xx , � xy [�] � 1 K
rxx [0] rxx [0]ryy [0] rxy [�] � lim � x[n] y[n � �]
K �� 2 K � 1 n � � K
• They are often used for convenience in
comparing and displaying • The autocorrelation sequence of a power
signal x[n] is given by
• Note: | � xx [�]| � 1 and | � xy [�]| � 1 K
independent of the range of values of x[n] 1
rxx [�] � lim � x[n]x[n � �]
and y[n] K �� 2 K � 1 n � � K
49 Copyright © 2005, S. K. Mitra 50 Copyright © 2005, S. K. Mitra

Correlation Computation for Correlation Computation for

Periodic Signals Periodic Signals
• The cross-correlation sequence for a pair of • Note: Both rxy
~ ~ [� ] and r~~ [� ] are also
xx
~ ~
periodic signals of period N, x[n] and y[n] , periodic signals with a period N
is defined as • The periodicity property of the
1 N �1 ~
~ ~ [� ] �
rxy N � n �0
x[n] y~[n � �] autocorrelation sequence can be exploited to
• The autocorrelation sequence of a periodic determine the period of a periodic signal
signal x~[n] of period N is given by that may have been corrupted by an additive
1 N �1 ~ random disturbance
N �n �0
~ ~ [� ] �
rxx x[n]x~[n � �]

51 Copyright © 2005, S. K. Mitra 52 Copyright © 2005, S. K. Mitra

Correlation Computation for Correlation Computation for

Periodic Signals Periodic Signals
• Let x~[n] be a periodic signal corrupted by • The autocorrelation of w[n] is given by
the random noise d[n] resulting in the signal rww[�] � M1 �nM��01 w[ n]w[n � �]
w[n] � x~[n] � d [n]
which is observed for 0 � n � M � 1 where � M1 �nM��01 ( x~[n] � d [n])( x~[ n � �] � d [n � �])
M �� N � M1 � nM��01 x~[ n] x~[n � �] � M1 � nM��01 d [n]d [n � �]

� M1 � nM��01 x~[n]d [n � �] � M1 �nM��01 d [n]x~[n � �]

� rxx
~ ~[� ] � r [� ] � r~ [� ] � r ~[� ]
dd xd dx
53 Copyright © 2005, S. K. Mitra 54 Copyright © 2005, S. K. Mitra

9
 Correlation Computation for Correlation Computation for
Periodic Signals Periodic Signals
• The autocorrelation rdd [�] of d[n] will show
• In the last equation on the previous slide, rxx
~ ~ [� ]
a peak at � = 0 with other samples having
is a periodic sequence with a period N and rapidly decreasing amplitudes with
hence will have peaks at � � 0, N , 2 N , ... increasing values of |� |
with the same amplitudes as � approaches M
• Hence, peaks of rww[�] for � > 0 are
• As x~[n] and d[n] are not correlated, samples essentially due to the peaks of rxx
~ ~ [� ] and can
of cross-correlation sequences rxd
~ [� ] and rdx~ [� ]
be used to determine whether x~[n] is a
are likely to be very small relative to the periodic sequence and also its period N if
amplitudes of rxx~ ~ [� ]
the peaks occur at periodic intervals
55 Copyright © 2005, S. K. Mitra 56 Copyright © 2005, S. K. Mitra

Correlation Computation of a
Correlation Computation of a
Periodic Signal Using MATLAB
Periodic Signal Using MATLAB 60
40

• Example - We determine the period of the 20

Amplitude
sinusoidal sequence x[n] � cos(0.25n) ,
0
-20

0 � n � 95 corrupted by an additive -40

-60
uniformly distributed random noise of -20 -10 0
Lag index
10 20

amplitude in the range [�0.5, 0.5] • As can be seen from the plot given above,
• Using Program 2_8 of text we arrive at the there is a strong peak at zero lag
plot of rww[�] shown on the next slide • However, there are distinct peaks at lags that
are multiples of 8 indicating the period of the
sinusoidal sequence to be 8 as expected
57 Copyright © 2005, S. K. Mitra 58 Copyright © 2005, S. K. Mitra

Correlation Computation of a
Periodic Signal Using MATLAB
• Figure below shows the plot of rdd [�]
8

6
Amplitude

-2
-20 -10 0 10 20
Lag index

• As can be seen rdd [�] shows a very strong

peak at only zero lag
59 Copyright © 2005, S. K. Mitra

10