Documente Academic
Documente Profesional
Documente Cultură
X (t, i )
X (t )
!!
!
L[]
(t )
!Y!
!
t
Our goal is to study the output process statistics in terms of the input
process statistics and the system function.
1A stochastic
Linear systems
Y (t ) = L[ X (t )]
Linear-Time Invariant
(LTI) systems
X (t )
h (t )
LTI system
Y (t ) = h(t ) X ( )d
(6-1)
= h( ) X (t )d . = X (t ) h(t )
Memoryless
system
Strict-sense
stationary output.
Wide-sense
stationary input
Memoryless
system
Need not be
stationary in
any sense.
X(t) stationary
Gaussian
Memoryless
system
Note: the filters are not random
LTI system
h(t)
Y (t )
wide-sense
stationary process.
(a)
X (t )
strict-sense
stationary process
LTI system
h(t)
(b)
X (t )
Gaussian
process (also
stationary)
Linear system
Y (t )
strict-sense
stationary process
Y (t )
Gaussian process
(also stationary)
(c)
5
Y (t ) = L{X (t )} L{X (t t0 )} = Y (t t0 )
(6-2)
(t )dt = 1
(t )
h(t ) = L{ (t )}
LTI
h(t )
h (t )
(6-3)
Impulse
response of
the system
t
Impulse
Impulse
response
X (t )
Y (t )
LTI
Y (t ) = h(t ) X ( )d
arbitrary
input
X (t ) = X ( ) (t )d
(6-4a)
(6-1)
= h( ) X (t )d
Y (t ) = L{ X (t )} = L{ X ( ) (t )d }
= L{ X ( ) (t )d }
By Linearity
= X ( ) L{ (t )}d
+
By Time-invariance
+
= X ( )h (t )d = h( ) X (t )d .
(6-4b)
7
$
E[Y (t)] = E & h( )X(t )d ') =
%
(
= E[X(t)] h(t)
h( )E[X(t )]d
(6-5)
function
(6-6)
Y = E[Y (t )] = X h( )d
h(u) h(v)R
RY ( ) =
( + v u)dudv
(6-7)
X(t) and Y(t) are jointly WSS and have I/O cross-correlation
RXY ( ) =
(6-8)
by
RY ( ) =
Proof:
To show that
we recall that
RY ( ) =
Y (t) =
h(u) h(v)R
( + v u)dudv
(6-10)
So,
RY ( ) = E[Y (t + )Y (t)] = E #$
(6-11)
See if you can prove the other relationships on the previous chart
on your own !. If not, see Papoulis.
10
Y = E[Yn ] = X
RY [n] =
hn .
(6-12)
n =
h h R
i
[n + i j ].
(6-13)
i = j =
RXY [n] =
h R
i
[n i ].
(6-14)
i =
RY [n] =
i =
RXY [n i ].
(6-15)
11
G ( f ) = g (t )e
j 2ft
dt ,
g (t ) = G ( f )e j 2ft df
(6-16)
RX ( )e
j 2 f
d ,
RX ( ) =
SX ( f )e j 2 f df
(6-18)
13
(6-19)
X ( ) = x(t )e jt dt ,
x (t )dt = 21
2
2
|
X
(
)
|
d = E.
(6-20)
X (t )
Energy in ( , + )
PILLAI
14
(6-21)
X T ( ) = T X (t )e jt dt
so that the energy per unit time, or the power spectral density (PSD)
2
(6-22)
| X T ( ) |2 1 T
j t
=
X (t )e dt
T
2T
2T
represents the power distribution associated with that realization based
on ( T, T ). Notice that (6-22) represents a random variable for every
and its ensemble average gives, the average power distribution
based on ( T, T ). Thus the power spectral density is given by
15
# | X T ( ) |2 $ 1 T T
j ( t1 t2 )
*
PT ( ) = E %
=
E
{
X
(
t
)
X
(
t
)}
e
dt1dt2
&
1
2
T T
' 2T
( 2T
1 T T
j ( t1 t2 )
=
R
(
t
,
t
)
e
dt1dt2
(6-23)
T T XX 1 2
2T
represents the power distribution of X(t) based on ( T, T ). For wide
sense stationary (WSS) processes, it is possible to further simplify
(6-23). Thus if X(t) is assumed to be WSS, then RXX (t1 , t2 ) = RXX (t1 t2 )
and (6-23) simplifies to
1
PT ( ) =
2T
j ( t1 t2 )
R
(
t
t
)
e
dt1dt2 .
T T XX 1 2
(6-24)
)
e
(2T | |)d
2 T XX
2T
2T
= 2T RXX ( )e j (1 2|T| )d 0
to be the power distribution of the WSS. process X(t) over
( T, T ). Finally letting T in (6-25), we obtain
(6-25)
PILLAI
16
S XX ( ) = lim PT ( ) = RXX ( )e j d 0
T
(6-26)
to be the power spectral density of the WSS process X(t). Notice that
"$$(')!!!!!!!!!!!!*.+!!!!!!!!!!-$$(.)0 (6-27)
i.e., the autocorrelation function and the power spectrum of a WSS
Process form a Fourier transform pair, a relation known as the
Wiener-Khinchin Theorem. From (6-27), the inverse formula gives
RXX ( ) = 21
j
S
(
)
e
d
XX
(6-28)
2
(
)
d
=
R
(0)
=
E{|
X
(t)
|
}= P
XX
XX
(6-29)
From (6-29), the area under S XX ( ) represents the total power of the
process X(t), and hence S XX ( ) truly represents the power
spectrum. (See the figure on the next chart).
PILLAI
17
S XX ( )
ai a RXX (ti t j ) = ai a 21
i =1 j =1
*
j
i =1 j =1
= 21
*
j
S XX ( )e
n
XX
( ) i =1 ai e
j ti
j ( ti t j )
d 0.
(6-30)
(6-31)
PILLAI
18
= RXX ( ) cos d
= 2 0 RXX ( ) cos d = S XX ( ) 0
(6-32)
PILLAI
19
2
"
S
(
f
)df
=
E
X
X
# (t)$% = RX (0)
(c) SX ( f ) = SX ( f )
(b)
(6-33)
20
[w/Hz]
Power spectral
density
Autocorrelation
function
Probability density function
21
X( ) =
xn e
j 2 n
xn =
n=
1/2
1/2
X( )e j 2 n d (6-34)
(6-35)
RX [k ]e
k =
j 2k
1/ 2
RX [k ] =
1 / 2
S X ( )e j 2k d
(6-36)
23
(b)
-1 / 2
[ ]
2
S X ( )d = E X n = RX [0] .
(6-37)
(c) S X ( ) = S X ( ),
(d) for any integer n, S X ( + n) = S X ( ).
24
f (t ) F ( ),
g (t ) G( )
(6-39)
Then
since
f (t ) g (t ) F ( )G( )
{ f (t) g(t)} =
f (t) g(t)e jt dt
(6-40)
(6-41)
PILLAI
25
{ f (t) g(t)}=
f ( )g(t )d e jt dt
(6-42)
f ( )e j d g(t )e j (t ) d (t )=F ( )G( ).
+
S XY ( ) = {RXX ( ) h* ( )} = S XX ( )H * ( )
(6-43)
and since
+
h ( )e
where
d =
+
h(t )e
j t
dt
= H * ( ),
H ( ) = h(t )e jt dt
(6-44)
(6-45)
SYY ( ) = {RYY ( )} = S XY ( )H ( ) = S XX ( ) | H ( ) |2 .
(6-46)
The above (6-46) is very useful: the PSD of the output signal
is the product of the PSD of the input and the magnitude
squared of the transfer function.
PILLAI
26
SYY ( ) = q | H ( ) |2
(6-48)
Notice that the output spectrum captures the system transfer function
characteristics entirely, and for rational systems (6-48) may be
used to determine the pole/zero locations of the underlying system.
PILLAI
27
SX(f)
H(f)
SY(f) =|H(f)|2Sx(f)
Frequency Domain
The above relationship is very useful in system design and analysis.
28
B/2
RXX ( ) = B / 2 S XX ( )e j d = q B / 2 e j d
sin( B / 2)
= qB
= qB sinc( B / 2)
( B / 2)
R ( )
(6-50)
XX
| H ( )|2
qB
B / 2
(a) LPF
B/2
PILLAI
29
Example: Let
Y (t ) =
t +T
2T t T
X ( )d
(6-51)
Y (t ) = h(t ) X ( )d = h(t ) X (t )
so that
where
SYY ( ) = S XX ( ) | H ( ) |2 .
+T
H ( ) = T
1
2T
e jt dt = sinc(T )
h (t )
1 / 2T
T
(6-52)
(6-53)
(6-54)
PILLAI
30
so that
SYY ( ) = S XX ( ) sinc 2 ( T ).
(6-55)
sinc2 ( T )
S XX ( )
SYY ( )
31
RXX ( ) =
rk ( kT ).
k =
S XX ( ) =
k =
rk e jT 0,
(6-56)
S XX ( ) = S XX ( + 2 / T )
so that S XX ( ) is a periodic function with period
2
2B =
.
T
(6-57)
(6-58)
PILLAI
32
)
e
d
(6-59)
B XX
2B
and
1 B
2
(6-60)
r0 = E{| X (nT ) | } =
S ( )d
B XX
2B
represents the total power of the discrete-time process X(nT). The
input-output relations for discrete-time system h(nT) translates into
and
where
S XY ( ) = S XX ( ) H * (e j )
(6-61)
SYY ( ) = S XX ( ) | H (e j ) |2
(6-62)
H (e ) =
n =
h( nT ) e j nT
(6-63)
PILLAI
33
(6-64)
so,
S XX ( ) =
S XX ( 0 ) + S XX ( + 0 )
+ S nn ( ).
2
(a)
(6-66)
S XX ( )
Smm ( )
(6-65)
(b)
S mm ( 0 )
34
Input
Output
Linear system
Deterministic signals:
Random signals:
(6-66)
35