Digital Signal Processing Reference
In-Depth Information
A reduced-rank algorithm must extract the most important features of the pro-
cessed data by performing dimensionality reduction. This mapping is carried out by
a
M
×
D
rank-reduction matrix
S
D
on the received data as given by
S
D
r
(i),
¯
r
(i)
=
(2.6)
where, in what follows, all
D
-dimensional quantities are denoted with a “bar”. The
resulting projected received vector
r
(i)
is the input to a beamformer represented by
¯
T
. The filter output is
the
D
×
1 vector
w
¯
=[¯
w
1
¯
w
2
...
w
D
]
¯
w
H
x(i)
¯
= ¯
r
(i).
¯
(2.7)
In order to design a reduced-rank beamformer
w
we consider the following opti-
¯
mization problem
minimize
E
¯
r
(i)
= ¯
2
w
H
w
H
R
¯
w
¯
(2.8)
w
H
subject to
¯
a
(θ
k
)
¯
=
1
.
The solution to the above problem is
R
−
1
¯
a
(θ
k
)
w
opt
=
¯
a
H
(θ
k
)
R
−
1
¯
a
(θ
k
)
¯
(
S
D
RS
D
)
−
1
S
D
a
(θ
k
)
=
a
H
S
D
(θ
k
)(
S
D
RS
D
)
−
1
S
D
a
(θ
k
)
,
(2.9)
where the reduced-rank covariance matrix is
R
r
H
(i)
S
D
RS
D
and the
=
E
[¯
r
(i)
¯
]=
S
D
a
(θ
k
)
. The above development shows
that the choice of
S
D
to perform dimensionality reduction on
r
(i)
is very impor-
tant, and can lead to an improved convergence and tracking performance over the
full-rank beamformer. A key problem with the full-rank and the reduced-rank beam-
formers described in (
2.5
) and (
2.9
), respectively, is that their performance is dete-
riorated when they employ the presumed array steering vector
a
p
(θ
k
)
.Inthesesit-
uations, it is fundamental to employ a robust technique that can mitigate the effects
of the mismatches between the actual and the presumed steering vector.
reduced-rank steering vector is
a
(θ
k
)
¯
=
2.3.2 Robust Adaptive LCMV Beamformers
An effective technique for robust beamforming is the use of diagonal loading strate-
gies [
7
,
9
,
12
,
13
]. In what follows, robust full-rank and reduced-rank LCMV beam-
forming designs are described. A general approach based on diagonal loading is
employed for both full-rank and reduced-rank designs.
Search WWH ::
Custom Search