Assignment 5 Sol

Assignment-5(Solutions)
Bayesian/ MMSE Estimation for MIMO/OFDM Wireless Communications
1. Linear estimate of channel vector h is, linear combination of observations, i.e.,

ĥ = c1 y(1) + c2 y(2) + ... + cN y(N ) = cT y
   
c1 y(1)
 c2   y(2) 
   
, where, c =  .  and y = 
 
 . 

 .   . 
cN y(N )
Mean Squared Error, 2
MSE = E ĥ − h

Minimum Mean Squared Error,

2
MMSE = minĥ E ĥ − h

For LMMSE, ĥ = cT y.
Hence, the minimization problem for LMMSE estimate is,
2
minĥ E ĥ − h ,

where ĥ = cT y.
Ans(c)
2. Considering multi-antenna channel problem,
y = Xh + v
where, y is Nx1 observation vector, X is NxM pilot matrix, h is Mx1 channel vector, v is
Nx1 noise vector, and N= number of pilot vectors, M= number of antennas. Also, given that
E{hi } = 0, E{|hi |2 } = σh2 , i.e.,
E{hhT } = σh2 I
Also,
Ryy = E{yyT } = σh2 XXT + σ 2 IM XM ,
Rhy = E{hyT } = σh2 XT
LMMSE estimate of channel vector h is,
ĥ = Rhy R−1
yy y
Substituting the values of Rhy and Ryy ,and Using the identity,
σh2 XT (σh2 XXT + σ 2 IM XM )−1 = σh2 (σh2 XT X + σ 2 IM XM )−1 XT
we obtain,
ĥ = σh2 XT (σh2 XXT + σ 2 IM XM )−1 y
= σh2 (σh2 XT X + σ 2 IM XM )−1 XT y
Ans(a)
1
3. LMMSE estimate,
ĥ = σh2 (σh2 XT X + σ 2 I)−1 XT y
For high Signal to Noise Ratio(SNR) scenario, σh2 XT X >> σ 2 I, thus, LMMSE estimate
reduces to,
ĥ = σh2 (σh2 XT X)−1 XT y
= (XT X)−1 XT y
For low Signal to Noise Ratio(SNR) scenario, σh2 XT X << σ 2 I, thus, LMMSE estimate reduces
to,
ĥ = σh2 (σ 2 I)−1 XT y
σh2 T
= X y
σ2
Ans(b)
4. The error covariance for the LMMSE estimate of h is,
E{(ĥ − h)(ĥ − h)T } = Rhh − Rhy R−1

yy Ryh .
From the solutions to the above problem Rhh , Rhy and Ryy can be written as,
Rhh = σh2 I,
Rhy = σh2 XT = (Ryh )T ,
Ryy = σh2 XXT + σ 2 I.
Substituting the values in the above equation, we obtain
E{(ĥ − h)(ĥ − h)T } = σh2 I − σh2 XT (σh2 XXT + σ 2 I)−1 σh2 X

= σh2 I − σh2 (σh2 XT X + σ 2 I)−1 XT σh2 X
= σh2 I − σh2 (σh2 XT X + σ 2 I)−1 (σh2 XT X + σ 2 I − σ 2 I)
= σ 2 σh2 (σh2 XT X + σ 2 I)−1
I −1
T
X X
= + 2 .
σ2 σh
Ans (c)
5. For a multi-antenna channel estimation scenario with N = 4 pilot vectors

1 1 2 1
x(1) = , x(2) = , x(3) = , x(4) = .
1 −2 1 −1
2
The corresponding pilot matrix X is given by,
x(1)T
     
1 1 2
x(2)T  1 −2 1
X=x(3)T  = 2 1  , y = 1 ,
    
x(4)T 1 −1 3
 
1 1
T 1 1 2 1  1 −2 7 0
X X=   =
1 −2 1 −1 2 1  0 7
1 −1
1
0
(σh2 XT X + σ 2 I)−1 = 15 1
0 15
 
2
T 1 1 2 1  1 8
X y=   = .
1 −2 1 −1 1   −2
3
1
Given data: σ 2 = −3 dB =⇒ 10 log σ 2 = −3 =⇒ σ 2 = 2 and σh2 = 0 dB =⇒ 10 log σ 2 =
0 =⇒ σ 2 = 1.
The LMMSE estimate of h is given by,
ĥ = σh2 (σh2 XT X + σ 2 I)−1 XT y

16/15
= .
−4/15
Ans (d)
6. The ML estimate of h for a multi-antenna channel estimation scenario is given by,
ĥ = (XT X)−1 XT y.
From the solution to above problem, we obtain

8/7
ĥ = .
−2/7
Ans (b)
7. The error covariance of the LMMSE estimate of h for a multi-antenna channel estimation
scenario is given by,
I −1
T
T X X
E{(ĥ − h)(ĥ − h) } = + 2 .
σ2 σh
From the solution to problem 5, we obtain
1

T 15 0
E{(ĥ − h)(ĥ − h) } = 1 .
0 15
Ans (a)
3
8. The error covariance of the LMMSE estimate of h for a multi-antenna channel estimation
scenario is given by,
I −1
T
T X X
E{(ĥ − h)(ĥ − h) } = + 2 .
σ2 σh
When σh2 XT X >> σ 2 I, the error covariance of the LMMSE estimate of h reduces to,
XT X −1

T
E{(ĥ − h)(ĥ − h) } =
σ2
1
0
= 14 1 .
0 14
Ans (a)
9. For a multi-antenna channel estimation scenario with N = 4 pilot vectors,

1 1 2 1
x(1) = , x(2) = , x(3) = , x(4) = .
1 −2 1 −1
The corresponding pilot matrix X is given by,
x(1)T
   
1 1
x(2)T  1 −2
X= x(3)T  = 2 1  .
  
x(4)T 1 1
The received output vectors are,

       
1 2 −2 1
y(1) = −1 , y(2) = −2 , y(3) = 1 , y(4) = 2  .
      
1 1 −2 −1
Therefore, the observation matrix Y is,
y(1)T
   
1 −1 1
y(2)T   2 −2 1 
Y= y(3)T  = −2 1 −2 .
  
y(4)T 1 2 −1
The MIMO system model is,

Y = XH + V.
Here, size of Y matrix is 4 x 3 and size of X matrix is 4 x 2. Therefore, size of H matrix or
size of the MIMO system is 2 x 3.
Ans (c)
10. The size of H matrix 2 X 3, i.e. M=2 and R=3 and the size of V matrix is 4 X 3. The
channel coefficients are IID Gaussian with dB variance σh2 = 0 dB = 1 and the noise samples
4
are also IID Gaussian with dB noise variance σ 2 = −3 dB = 12 . Therefore, the covariance
matrices RHH of the channel matrix and RV V of the noise matrix, respectively are given by,
RHH = E{HHT }
= Rσh2 I2X2
= 3I2X2
RV V = E{VVT }
= Rσ 2 I4X4
3
= I4X4 .
2
Ans (a)

Assignment 5 Sol

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Assignment 5 Sol

Uploaded by

Copyright:

Available Formats

Assignment-5(Solutions)

Bayesian/ MMSE Estimation for MIMO/OFDM Wireless Communications

1. Linear estimate of channel vector h is, linear combination of observations, i.e.,

Minimum Mean Squared Error,

4. The error covariance for the LMMSE estimate of h is,

E{(ĥ − h)(ĥ − h)T } = Rhh − Rhy R−1

Substituting the values in the above equation, we obtain

E{(ĥ − h)(ĥ − h)T } = σh2 I − σh2 XT (σh2 XXT + σ 2 I)−1 σh2 X

5. For a multi-antenna channel estimation scenario with N = 4 pilot vectors

The corresponding pilot matrix X is given by,

ĥ = σh2 (σh2 XT X + σ 2 I)−1 XT y

6. The ML estimate of h for a multi-antenna channel estimation scenario is given by,

From the solution to above problem, we obtain

9. For a multi-antenna channel estimation scenario with N = 4 pilot vectors,

The corresponding pilot matrix X is given by,

The received output vectors are,

Therefore, the observation matrix Y is,

The MIMO system model is,

You might also like