Digital Signal Processing Reference
In-Depth Information
Adaptive Weighted Median
1
0.8
0.6
µ
= 0.04
0.4
µ
= 0.15
0.2
0
0
50
100
150
200
250
300
(a)
Adaptive Weighted Myriad
1
0.8
0.6
µ
= 0.8
0.4
µ
= 3
0.2
0
0
50
100
150
200
250
300
(b)
FIGURE 5.14
Comparison of convergence rate of the optimal weighted median and the optimal weighted
myriad: (a) optimal weighted median at
µ
=
0
.
04
,
0
.
15, (b) optimal weighted myriad at
µ
=
0
.
8
,
3.
applications like DSL, it has been shown that impulsive noise is prevalent,
32
where inevitably, CMA blind equalization using FIR structure collapses. Here
we describe a novel real-valued blind equalization algorithm, which is the
combination of the constant modulus criterion and the weighted myriad fil-
ter structure. Using myriad filters, we should expect a very close to linear
performance when we set the linear parameter
K
to be far larger than the
data samples. When the noise contains impulses, by reducing
K
to a suitable
level, we can manage to remove their influence greatly without losing the
capability of keeping the communication eye open.
Consider a pulse amplitude modulation (PAM) communication system;
signal and channel are all real. The constant modulus cost function is defined
as follows:
1
4
E
=
2
2
J
(
W
,K
)
{
(
|
Y
(
n
)
|
−
R
2
)
}
,
(5.86)
where
4
E
|
S
(
n
)
|
R
2
=
E
|
S
(
n
)
|
2
Search WWH ::
Custom Search