Digital Signal Processing Reference
In-Depth Information
V
†
λ
Υ
1
1
0
Υ
1
Υ
1
V
†
Υ
1
=
U
Υ
1
=
U
Υ
1
.
(4.45)
Υ
1
λ
Υ
1
2
0
Then (
4.44
) becomes
Υ
1
Υ
1
V
†
2
|
HU
1
|
=
|
λ
Υ
1
)
+|
λ
Υ
1
2
2
h
11
|
2
h
21
|
2
2
h
12
|
2
h
22
|
2
|
(
|
+|
|
(
|
+|
)
Υ
v
h
|
2
1
|
≥
L
L
(4.46)
where
h
11
h
12
HU
Υ
1
=
.
(4.47)
h
21
h
22
Since the unitary matrix
U
does not change the distribution of
H
, each element
Υ
1
Υ
1
, i.e.,
h
ij
, is also a Gaussian distributed random variable with mean 0 and
variance 1. As a result, (
4.43
) can be written as
of
HU
|
H
P
(
d
→
d
)
exp
4
L
1
2
(
|
λ
Υ
1
1
2
h
11
|
2
h
21
|
2
≤
−
|
cos
θ
hg
|
|
(
|
+|
)
2
+|
λ
Υ
1
2
2
h
12
|
2
h
22
|
2
|
(
|
+|
))
|
γ
1
+
γ
2
|
.
(4.48)
Further, we have
P
(
d
→
d
)
E
exp
4
L
(
|
1
2
(
|
λ
Υ
1
1
2
h
11
|
2
h
21
|
2
≤
−
cos
θ
hg
|
|
(
|
+|
)
+|
λ
Υ
1
2
2
h
12
|
2
h
22
|
2
2
|
(
|
+|
))
|
γ
1
+
γ
2
|
)
E
E
exp
4
L
(
|
(
|
λ
Υ
1
1
1
2
2
h
11
|
2
h
21
|
2
=
−
cos
θ
hg
|
|
(
|
+|
)
hg
+|
λ
Υ
1
2
2
h
12
|
2
h
22
|
2
2
1
|
(
|
+|
))
|
γ
1
+
γ
2
|
)
|
θ
⎡
⎤
1
⎣
⎦
≤
E
(4.49)
j
=
1
[
|
λ
Υ
1
j
+
(
8
L
|
1
2
2
2
2
1
cos
θ
hg
|
|
|
γ
1
+
γ
2
|
)
]
At high SNRs, one can neglect the one in the denominator and get
|·|
γ
1
+
γ
2
|
)
−
4
E
8
L
−
4
2
1
1
(
|
λ
Υ
1
P
(
d
→
d
)
≤
.
(4.50)
j
1
8
|
cos
θ
hg
|
j
=
Search WWH ::
Custom Search