Digital Signal Processing Reference
In-Depth Information
7.5 Adaptive Beamforming versus Differential Arrays
The linear system given in (7.3) can be generalized to
Ψ
N+1
h
(
ω
)=
a
N+1
(
ω
)
,
(7.40)
where now
11 1
···
1
01 2
··· M −
1
01 2
2
···
(
M −
1)
2
. . .
.
.
.
.
012
N
···
(
M −
1)
N
Ψ
N+1
=
(7.41)
is the constraint matrix of size (
N
+1)
×M
having the Vandermonde structure,
M
is the number of microphones,
T
h
(
ω
)=
H
1
(
ω
)
H
2
(
ω
)
···H
M
(
ω
)
(7.42)
is a filter of length
M
, and the vector
a
N+1
(
ω
) was already defined in (7.5).
In adaptive beamforming [2], we minimize the residual noise at the beam-
former output subject to the constraints summarized in (7.40). Mathemati-
cally, this is equivalent to
h(ω)
h
H
(
ω
)
Φ
v
(
ω
)
h
(
ω
) subject to
Ψ
N+1
h
(
ω
)=
a
N+1
(
ω
)
.
(7.43)
min
We easily deduce that the solution is the LCMV filter [2], [3]:
−1
−1
v
(
ω
)
Ψ
N+1
−1
v
(
ω
)
Ψ
N+1
h
LCMV
(
ω
)=
Φ
Ψ
N+1
Φ
a
N+1
(
ω
)
.
(7.44)
v
(
ω
)
Ψ
N+1
in (7.44) to be full rank,
we must have
N
+1
≤ M
, which is the same condition to design a differential
array of order
N
.
For
M
=
N
+ 1, we easily deduce from (7.44) that
−1
We observe that for the matrix
Ψ
N+1
Φ
−1
h
LCMV
(
ω
)=
Ψ
N+1
a
N+1
(
ω
)
,
(7.45)
which corresponds exactly to the filter of an
N
th-order DMA or the solution
of (7.40).
For
M>N
+ 1 and spatially white noise, (7.44) becomes
−1
h
LCMV
(
ω
)=
Ψ
N+1
Ψ
N+1
Ψ
N+1
a
N+1
(
ω
)
,
(7.46)
which corresponds to the minimum-norm solution of (7.40). This shows that
the LCMV filter is fundamentally related to the filter of an
N
th-order DMA.
Search WWH ::
Custom Search