Digital Signal Processing Reference
In-Depth Information
Choose
d
mq
's to minimize
J
. We must have
∂
J
=,
0
(2.53)
∗
∂
d
mq
ˆ
d
=
mq
d
mq
which leads to
Q
P
−
1
T
−
1
T
−
1
∑∑ ∑
()()
′
q
=
∑
d
( )
=
()()
j
unune
α−
α
n
j
nu ne
−
α
n
m
m
y
.
m
(2.54)
mq
′′
q
′
q
q
′= ′=
1
m
0
n
=
0
n
0
=
:
g
mq
Define
1
1
1
1
e
−
j
α
e
−
j
α
L
1
1
V
:=
,
(2.55)
−
j
α
−
j
α
L
1
e
e
P
−
1
P
−
1
T
H
T
T
H
H
Dd d
:= ,, , :=, ,
D
D
D
,
(2.56)
m
m
1
mQ
0
P
−
1
T
H
:=
()
,,
()
, :=, ,
T
T
H
H
Hh
l
h
l
H
H
H
,
(2.57)
l
1
Q
0
L
(
)
⊗.
C:=
diag c
,,
VI
c
P
(2.58)
0
−
NQ
By the definition of
d
mq
in (2.56), it then follows that
CH =.
(2.59)
It is shown in [57] that if
P
≥
L
+ 1,
rank
(
C
) =
NQ
(
L
+ 1). Hence, we can determine the
h
q
(
l
)'s uniquely by using the estimates of
d
mq
's.
Define D
ˆ
as in (2.56) with
d
mq
replaced with
ˆ
mq
, and similarly define G as in (2.56)
with
d
mq
replaced with
g
mq
. Then (2.54) leads to
Ψ
( )
=
ˆ
DG,
(2.60)
I
N
where the entries of the
PQ
×
PQ
matrix Ψ are (
m
,
m
′ = 0, 1;
…
,
P
- 1,
q
,
q
′ = 1, 2,
…
,
Q
)
Search WWH ::
Custom Search