Digital Signal Processing Reference
In-Depth Information
x
ð
n
Þ¼
X
M
h
ð
k
Þ
y
ð
n
k
Þ
ð
3
:
46
Þ
k
¼
0
Often it is convenient to write the above equation in matrix form as follows:
x
ð
n
Þ¼
hy
T
ð
n
Þ¼
y
ð
n
Þ
h
T
ð
3
:
47
Þ
where
h
¼½
h
ð
0
Þ
h
ð
1
Þ
h
ð
2
Þ;
...
;
h
ð
M
Þ;
which is the impulse response vector, and
y
ð
n
Þ¼½
y
ð
n
Þ
y
ð
n
1
Þ
y
ð
n
2
Þ;
...
;
y
ð
n
M
Þ;
which is the observed signal vector at the time instant n. The error is given by:
e
ð
n
Þ¼
d
ð
n
Þ
x
ð
n
Þ
ð
3
:
48
Þ
and the mean-squared error (MSE) is given by:
e
mse
¼Ef½
e
ð
n
Þ
2
g¼Ef½
d
ð
n
Þ
x
ð
n
Þ
2
g¼Ef½
d
ð
n
Þ
X
M
h
k
y
ð
n
k
Þ
2
g
3
:
49
Þ
k
¼
0
To minimize e
mse
it is necessary that:
o
e
mse
oh
j
¼
0
8
j
ð
3
:
50
Þ
where for simplicity, the designation h
j
= h(j) is made. From Eqs.
3.47
,
3.49
, and
3.50
it is possible to write:
o
e
mse
oh
j
¼E
2e
ð
n
Þ:
o
e
ð
n
Þ
oh
j
ð
3
:
51
Þ
¼
2
Ef
e
ð
n
Þ
y
ð
n
j
Þg
¼
2
Ef
y
ð
n
j
Þ½
d
ð
n
Þ
y
ð
n
Þ
h
T
g ¼
0;
j
¼
0
;
1
;
...
;
M
Now Eq.
3.51
can be written in matrix form as follows:
2
Ef
y
T
ð
n
Þ½
d
ð
n
Þ
y
ð
n
Þ
h
T
g ¼
0
)
Ef
y
T
ð
n
Þ
d
ð
n
Þg Ef
y
T
ð
n
Þ
y
ð
n
Þ
h
T
g ¼
0
) R
yd
¼
R
yy
h
T
where R
yy
is the observed signal autocorrelation, while R
yd
is the observed signal
cross-correlated with the desired signal. Hence, the optimal filter coefficients (in
the MSE sense) are given by:
Search WWH ::
Custom Search