Digital Signal Processing Reference
In-Depth Information
P.2
In general, a
cross-correlation function is neither even nor odd
. The following
relation holds:
R
XY
ðtÞ ¼ R
YX
ðtÞ:
(6.113)
From the definition of a cross-correlation function (
6.94
)-(
6.95
), we write
R
XY
ðtÞ ¼ E XðtÞYðttÞ
g
:
f
(6.114)
Introducing
t
0
¼ tt;
(6.115)
into (
6.114
), we can write:
R
XY
ðtÞ ¼ E Yðt
0
ÞXðt
0
þ tÞ
f
g ¼ R
YX
ðtÞ:
(6.116)
P.3
This property establishes the upper limit for a cross correlation function as a
geometric mean of the corresponding autocorrelation functions at the origin:
j
p
j
R
XY
ðtÞ
R
XX
ð
0
ÞR
YY
ð
0
Þ
:
(6.117)
For any two random variables, we can write:
2
EfX
2
gEfY
2
½
EfXYg
g:
(6.118)
Then we have:
ðtÞ
E Y
2
2
E X
2
½
E XðtÞYðt þ tÞ
f
g
ðt þ tÞ
(6.119)
resulting in the desired result (
6.117
).
P.4
This property establishes another upper limit for a cross-correlation function as
an arithmetic mean of the corresponding autocorrelation functions in the origin.
This is a stronger limit than that of (
6.117
) because a geometric mean of two
positive numbers cannot exceed an arithmetic mean of the same numbers:
1
2
R
XX
ð
0
ÞþR
YY
ð
0
Þ
j
R
XY
ðtÞ
j
½
:
(6.120)
To establish this property, we write:
2
¼ X
2
ðtÞþY
2
½
XðtÞYðt þ tÞ
ðtÞ
2
XðtÞYðt þ tÞ
0
:
(6.121)
Search WWH ::
Custom Search