Digital Signal Processing Reference
In-Depth Information
or alternatively
8
<
9
=
1
T=
2
ð
1
T
e
jot
d
t:
S
XX
ðoÞ¼
lim
T!1
R
XX
ðt; t þ tÞ
d
t
(7.27)
:
;
1
T=
2
The expression in the brackets presents a time average of the autocorrelation
function of the process
X
(
t
)
T=
2
ð
1
T
AR
XX
ðt; t þ tÞ
f
g ¼
lim
T!1
R
XX
ðt; t þ tÞ
d
t:
(7.28)
T=
2
If we suppose that the process is at least WS stationary, its autocorrelation function
does not depend on absolute time but only of a time difference
t
, resulting in:
T=
2
ð
1
T
AR
XX
ðt; t þ tÞ
g
¼
lim
T!1
R
XX
ðtÞ
d
t ¼ R
XX
ðtÞ:
f
(7.29)
T=
2
Placing (
7.29
) into (
7.27
), we finally get:
1
R
XX
ðtÞ
e
jot
d
t;
S
XX
ðoÞ¼
(7.30)
1
or equivalently
1
1
2
p
S
XX
ðoÞ
e
jot
d
t:
R
XX
ðtÞ¼
(7.31)
1
The obtained expressions (
7.30
) and (
7.31
) are known as the
Wiener-Khinchin
theorem
, which states that a
PSD and a autocorrelation function of a process, which
is at least WS stationary, are the Fourier transform pair.
The importance of this theorem is that one can obtain a spectral characteristic
(i.e., the PSD of a random process as the Fourier transform of an autocorrelation
function), which itself is a deterministic function that represents the process.
Consequently, the Wiener-Khinchin theorem solves all problems related to the
application of the Fourier transform to a random process, mentioned in Sect.
7.1.2
.
Example 7.1.1
Find a PSD for the process considered in Example 6.5.2, where the
autocorrelation is obtained in the expression (
6.77
).
Search WWH ::
Custom Search