LCMV(线性约束最小方差)实际上是MVDR(最小方差无畸变响应)的一种扩展,将MVDR里只对期望方向做约束,变为也对指定的干扰方向加约束。
依然使用MVDR严谨推导里的模型,空间中存在期望信号 x d ( t ) x_{d}(t) xd(t)其导向矢量为 a ( θ d ) a\left(\theta_{d}\right) a(θd),以及M-1个干扰信号 x j ( t ) x_{j}(t) xj(t)其导向矢量为 a ( θ j ) a\left(\theta_{j}\right) a(θj),阵列个数为N
LCMV要求解的问题可以写为:
{
min
W
H
R
W
s.t.A
H
(
θ
)
W
=
g
\left\{\begin{array}{c} \min W^{H} R W \\ \text {s.t.A}^{H}(\theta) W=g \end{array}\right.
{minWHRWs.t.AH(θ)W=g
其中
A
(
θ
)
=
[
a
(
θ
d
)
,
a
(
θ
j
)
,
…
.
a
(
θ
M
−
1
)
]
N
×
M
A(\theta)=\left[a\left(\theta_{d}\right), a\left(\theta_{j}\right), \ldots . a\left(\theta_{M-1}\right)\right]_{N \times M}
A(θ)=[a(θd),a(θj),….a(θM−1)]N×M
g
=
[
1
0
⋯
0
]
M
×
1
g=\left[\begin{array}{c}1 \\ 0 \\ \cdots \\ 0\end{array}\right]_{M \times 1}
g=
10⋯0
M×1
构造拉格朗日方程:
L
(
ω
)
=
ω
H
R
ω
−
(
ω
H
A
(
θ
)
−
g
)
λ
⃗
L(\omega)=\omega^{\mathrm{H}} \boldsymbol{R} \omega-\left(\omega^{\mathrm{H}} \boldsymbol{A(\theta)}-g\right)\vec{\lambda}
L(ω)=ωHRω−(ωHA(θ)−g)λ
得:
∂
L
(
w
)
∂
w
H
=
R
w
−
A
(
θ
)
λ
⃗
=
0
w
o
p
t
=
R
−
1
A
(
θ
)
λ
⃗
\begin{array}{l} \frac{\partial L(w)}{\partial w^{H}}=R w- A(\theta) \vec{\lambda}=0 \\ w_{o p t}=R^{-1} A(\theta) \vec{\lambda} \end{array}
∂wH∂L(w)=Rw−A(θ)λ=0wopt=R−1A(θ)λ
两端同乘约束:
A
H
(
θ
)
w
o
p
t
=
A
H
(
θ
)
R
−
1
A
(
θ
)
λ
⃗
g
=
A
H
(
θ
)
R
−
1
A
(
θ
)
λ
⃗
λ
⃗
=
(
A
H
(
θ
)
R
−
1
A
(
θ
)
)
−
1
g
\begin{array}{l} A^{H}(\theta) w_{o p t}=A^{H}(\theta) R^{-1} A(\theta) \vec{\lambda} \\ g=A^{H}(\theta) R^{-1} A(\theta) \vec{\lambda} \\ \vec{\lambda}=\left(A^{H}(\theta) R^{-1} A(\theta)\right)^{-1} g \end{array}
AH(θ)wopt=AH(θ)R−1A(θ)λg=AH(θ)R−1A(θ)λλ=(AH(θ)R−1A(θ))−1g
将
λ
\lambda
λ带入
w
o
p
t
w_{opt}
wopt得
w
o
p
t
=
R
−
1
A
(
θ
)
(
A
H
(
θ
)
R
−
1
A
(
θ
)
)
−
1
g
w_{o p t}=R^{-1} A(\theta)\left(A^{H}(\theta) R^{-1} A(\theta)\right)^{-1} g
wopt=R−1A(θ)(AH(θ)R−1A(θ))−1g
推导完毕。