Modified Whitening Rotation based Joint Semi-blind Channel and Data Estimation Scheme for Rayleigh Flat Fading MIMO channels

ACEEE Int. J. on Communications, Vol. 03, No. 01, March 2012

Modified Whitening Rotation based Joint Semi-blind
Channel and Data Estimation Scheme for Rayleigh
Flat Fading MIMO channels
Jaymin Bhalani A. I. Trivedi
Dept. of Electronics & Communication Engineering, Dept. of Electrical Engineering,
Faculty of Technology & Engineering, Faculty of Technology & Engineering,
Charotar University of Science and Technology The M.S. University of Baroda, Gujarat, India.
Changa, Gujarat, India. E-mail: ai.trived@gmail.com
E-mail: jayminbhalani.ec@charusat.ac.in

Abstract— In this paper, we propose a novel joint semi-blind effects caused by the scatterers. Channel state information
channel and data estimation technique based on Whitening (CSI) provides key information for the operation of MIMO
Rotation (WR) method for Rayleigh flat fading Multiple Input wireless communication systems and hence need to be esti-
Multiple output (MIMO) channel using different receiver
mated accurately. Many channel estimation algorithms have
antennas combinations. Here we divide newly proposed
been developed in recent years. In the literature [11-14], MIMO
technique in three steps. In the first step, we use conventional
Whitening Rotation based semi-blind channel estimation channel estimation methods can be classified into three
technique, where MIMO channel matrix H can be decomposed classes: training based, blind and semi-blind. For pure train-
as H=WQ H . Whitening matrix W can be estimated blindly ing based scheme, a long training is necessary in order to
using second order statistical information of received data obtain a reliable channel estimate, which considerably re-
and unitary rotation matrix Q can be estimated exclusively duces system bandwidth efficiency. In Blind methods, no
using Orthogonal Pilot Maximum Likelihood (OPML) training symbols are used and channel state information is
algorithm. In the second step, data symbols can be estimated acquired by relaying on the received Signal statistics [17-20],
using estimated channel H and received output data by
which achieves high system throughput requiring high com-
applying maximum likelihood data estimation method.
putational complexity. Semi-blind channel estimation ap-
Finally in the third step, Q can be re-estimated as a Q new
using OPML algorithm by considering estimated blind data proaches as a combination of the two aforementioned proce-
symbols itself as a pilot symbols for more statistical dures [21-23], with few training symbols along with blind
information of unitary matrix and perform final channel statistical information, such techniques can solve the con-
estimation H final=W Q new H . Simulation results are presented vergence problems and high complexity associated with blind
under 4-PSK data modulation scheme for two transmitters estimators. Extensive work has been done later by Slock et.
and different combinations of receiver antennas to support al. [3-4], where several semi-blind techniques have been re-
proposed novel technique and they demonstrate improved BER ported. Whitening Rotation (WR) based semi blind technique
performance compared to conventional WR based optimal
with Orthogonal Pilot Maximum Likelihood (OPML) [5-8] has
technique and Rotation Optimization Maximum Likelihood
shown very good performance compare to other sub-optimal
(ROML) based suboptimal semi-blind channel estimation
technique. techniques and training based channel estimation techniques.
In WR method, channel matrix H is decomposed as whiten-
Keywords- Multiple Input Multiple Output, Orthogonal Pilot ing matrix W and unitary rotation matrix Q. Whitening matrix
Maximum Likelihood technique, Whitening Rotation based W is estimated using received output data blindly and Q is
Semi Blind Channel Estimation , Rotation Optimization estimated using orthogonal pilot maximum likelihood (OPML)
Maximum Likelihood technique algorithm.
Here we have developed a novel joint semi-blind channel
I. INTRODUCTION and data estimation technique which is described by three
A Multiple Input Multiple Output (MIMO) communication basic steps given below
system uses multiple antennas at the transmitter and receiver Step 1: Initial channel estimation is performed using
to achieve numerous advantages. Traditionally, antenna Whitening Rotation (WR) based semi-blind
arrays have been used at the transmitter and the receiver to channel estimation technique as H=WQ H.
achieve array gain, which increases the output SNR of the Step 2: Given channel knowledge (estimate) and received
system. More recently, ways of using multiple antennas has output, perform data estimation using maximum
been discovered to achieve diversity and multiplexing gain likelihood method.
by exploiting the once negative effect of multipath. Under Step 3: Estimate unitary rotation matrix Q new again by
suitable conditions, i.e. a scatter rich environment, the channel Considering estimated blind data symbols itself
paths between the different transmit and receive antennas as pilot symbols and perform final channel
can be treated as independent channels due to the multipath estimation as H final=W Q new H.
© 2012 ACEEE 35
DOI: 01.IJCOM.3.1.514


This novel technique performs better than conventional WR
unitary i.e. Q H Q  QQ H  I As shown in [5], the matrix
based optimal and rotation optimization ML (ROML) based
suboptimal semi-blind channel estimation techniques. The W can be estimated from the covariance of received data
remainder of this paper is organized as follows. The second alone. We therefore employ the pilot information to exclu-
section describes the system model. The estimation algorithms
sively estimate the rotation matrix Q . This semi-blind esti-
with proposed techniques are presented in section 3,
simulation results and discussion provides in section 4 and mation procedure is termed as a Whitening-Rotation (WR)
section 5 concludes this paper. scheme. Let the Singular Value Decomposition (SVD) of H be
given as P  Q H . A possible choice for W is given by
II. SYSTEM MODEL
W  P  and we assume this speciûc choice in the rest of
Consider a ûat fading MIMO channel matrix
the work. We present next a list of potential assumptions
H  c rt where t is the number of transmit antennas and r is which are employed as appropriate in subsequent parts of
the number of receive antennas in the system, and each the work.
hij represents the ûat-fading channel coefficient between the Assumption A. W  c rt is perfectly known at the output.

th
i th receiver and j transmitter. Denoting the complex
Assumption B. Xp  c t L is orthogonal i.e.

received data by Y  c r 1 , the equivalent base-band system XpXp H   s2 LI tt
can be modeled as Assumption A is reasonable if we assume the transmission
of a long data stream (N ) from which W can be
Y ( k )  HX ( k )   ( k ) (1)
estimated with considerable accuracy and Assumption B can
be easily achieved by using an integer orthogonal structure
 h11 h12  h1t  such as the Hadamard matrix.
h h22  h2t   
H   21  Q : c r  L  S , where Q the constrained ML estimator of is
Where      
  Q and S is the manifold of unitary matrices, is obtained by
 hr1 hr 2  hrt 
minimizing the likelihood
k represents the time instant, X  c t1 is the complex trans- 2
Yp  WQ H Xp Such that QQ H  I (2)
mitted symbol vector.  is additive white Gaussian noise such
2
that { ( k ) (l )}   ( k , l ) n I where  ( k , l )  1 if M  W H YpXp H We then have the following result for the

k  1 and 0 otherwise. Also, the sources are assumed to be constrained estimation of Q . Under both assumption Q the
spatially and temporally independent with identical source
constrained OPML estimate of Q that minimizes the cost
power  s2 i.e. { X ( k ) X (l )}   ( k , l ) s2 I . The signal to
function in (2) is given by
 s2  H H
noise ratio (SNR) of operation is deûned as SNR 2 As- Q = VM U M Where, U M  M VM  SVD ( M ) (3)
n
This technique has been proposed and proved in [5] since
sume that the channel has been used for a total of N symbol this procedure employs assumption B. (orthogonal pilot),
transmissions. Out of these N transmissions, the first L sym- which termed as the OPML estimator. The above expression
bols are known training symbols Xp and reaming N - L are (3) thus yields a closed form expression for the computation
blind data symbols Xb , Yp and Yb are their respective outputs. 
of Q , the ML estimate of Q . The channel matrix H is then
III. CHANNEL ESTIMATION TECHNIQUES
estimated
A). Whitening Rotation based semi-blind channel estimation  
H
as
Now consider a MIMO channel H  c rt which has at H WQ .
= (4)

least as many receive antennas as transmit antennas i.e. B). Rotation Optimization Maximum likelihood (ROML)
r  t .Then, the channel matrix H can be decomposed as based semi-blind channel estimation
To avoid the complexity involved in the full computation
H  WQ H where W  c rt is also known as the whitening of the optimal ML solution, we propose a simplistic ROML
matrix and Q  c tt , termed as the rotation matrix, that is procedure for the sub-optimal estimation of Q , thus trading
© 2012 ACEEE 36
DOI: 01.IJCOM.3.1.514


complexity for optimality. The ûrst step of construct modified IV. SIMULATION RESULTS
cost function
Extensive computer simulations have been conducted to
2
demonstrate the performance of proposed novel technique
m in W Y p -Q H X P compared with conventional WR based optimal and ROML
Q Where QQ H  I (5) based suboptimal techniques. In simulation scenarios the 4-
PSK data modulation scheme is used and flat fading Rayleigh
MIMO channels H are generated. We consider alamouti coded
Y  W Yp Is the whitening pre-equalized data. Several choices 2×N MIMO (where N= 4, 6, 8 receiver antennas) systems
with 4 orthogonal pilots and first 100 blind data symbols
can then be considered for the pre-equalization ûlter W . A ro- among 20000 pair transmitted symbols used. Result shows in
bust MMSE pre-ûlter is given as figures depicts that novel technique outperforms others by 2
dB to 3 dB performance improvements due to more statistical
2 H 2 H 2 1
W MMSE   s W ( s WW   n I ) (6)
information of unitary matrix Q by using whole burst of
estimated data itself as pilot symbols in OPML algorithm. It
 achieves nearby performance compared to perfect CSI
Defining D W YpXp H , the cost minimizing Q for the (channel state information). Now in first simulation setup
modifying cost is given as shown in figure 1, we have used 2 transmitter and 4 receiver
 H
antennas. At 2 dB SNR, Bit error rate (BER) is 0.0853 for
Q  VDU D Where U D S DVDH  SVD ( D ) (7) ROML, 0.0619 for WR, 0.0242 for new technique and 0.0213
for perfect CSI.
 
The channel matrix H is then estimated as H = W Q H.

C). A novel joint semi-blind channel and data estimation
technique (proposed new-WR technique)
Initial channel estimation is accomplished by WR based

technique. Now based on that channel estimate H , perform
data estimation Xbest as

 H
X best  H Yb (8)
Further in Xbest itself consider as pilot symbols, Yb is received
output and again perform OPML algorithm as Fig. 1 BER Vs SNR for diff. channel estimation techniques using 4
pilot symbols and 100 Blind symbols using 2 transmitter and 4
2
Yb  WQnew H Xbest Such that QnewQnew H  I
(9)
receiver antennas for 4 PSK modulation scheme
In the second simulation setup shown in figure 2, we have
And Mb is denoted as used 2 transmitter and 6 receiver antennas. At 2 dB SNR, Bit
error rate (BER) is 0.0303 for ROML, 0.0207 for WR, 0.0069 for
(10) new technique and 0.0057 for perfect CSI.
Mb  W H YbXbest H

Under both assumption Q new the constrained OPML estimate

of Qnew that minimizes the cost function in (9) is given by
 H H
Q new = VMb U Mb Where, U Mb  Mb VMb  SVD ( Mb )
(11)
So final channel estimation Hfinal is given by
 
H final
= W Q new H. (12)

Fig. 2 BER Vs SNR for diff. channel estimation techniques using 4
pilot symbols and 100 Blind symbols using 2 transmitter and 6
receiver antennas for 4 PSK modulation scheme

© 2012 ACEEE 37
DOI: 01.IJCOM.3.1. 514


Third simulation setup shown in figure 3, in that we have REFERENCES
used 2 transmitter and 8 receiver antennas. At 2 dB SNR, Bit
[1] J.G.Proakis, Digital Communications, McGraw-Hill Higher
error rate (BER) is 0.0110 for ROML, 0.0079 for WR, 0.0016 for Education, New York, 2001.
new technique and 0.0013 for perfect CSI. [2] D Pal, “Fractionally spaced semi-blind equalization of wireless
channels,” The Twenty-Sixth Asilomar Conference, 1992, vol. 2,
pp. 642–645.
[3] E. De Carvalho and D. T. M. Slock, “Asymptotic performance
of ML methods for semi-blind channel estimation,” Thirty-First
Asilomar Conf., 1998, pp. 1624–1628.
[4] A.Medles, D.T.M. Slock, and E.D.Carvalho, “Linear prediction
based semi-blind estimation of MIMO FIR channels,” Third
IEEE SPAWC, Taiwan, pp.58-61, 2001.
[5] A. K. Jagannatham and B. D. Rao, “A semi-blind technique for
MIMO Channel matrix estimation,” in Proc. of IEEE Workshop
on Signal Processing Advances in Wireless Communications
(SPAWC 2003), Rome, Italy, pp no. 304-308.
[6] Xia Liu, Feng Wang and Marek E. Bialkowski “Investigation
into a Whitening-Rotation-Based Semi-blind MIMO Channel
Estimation for Correlated” International Conference on Signal
Processing and Communication Systems, ICSPCS 2008. pp no-1-
Fig. 3 BER Vs SNR for diff. channel estimation techniques using 4 4.
pilot symbols and 100 Blind symbols using 2 transmitter and 8 [7] Qingwu Zhang, Wei-Ping Zhu, Qingmin Meng, ‘Whitening-
receiver antennas for 4 PSK modulation scheme rotation-based semi-blind estimation of MIMO FIR channels’
Further with respect to increases in receiver antennas .International Conference on Wireless Communications & Signal
combinations, BER improves 1.5 dB to 2 dB respectively that Processing, WCSP 2009, pp. 1-4.
shows in figure 4. [8] Feng Wan, Wei-Ping Zhu and M.N.S. Swamy ‘Perturbation
Analysis of Whitening-Rotation-based Semi-Blind MIMO Channel
Estimation’ IEEE International Midwest Symposium on Circuits
and Systems, MWSCAS ’09,2009 pp 240-243.
[9] F. Wan, W.-P. Zhu, and M. N. S. Swamy, “A semi-blind channel
estimation approach for MIMO-OFDM systems”, IEEE Trans.
on Signal Processing, vol. 56, no. 7, pp. 2821–2834, 2008..
[10] A. K. Jagannatham and B. D. Rao, “Whitening-rotation-based
semiblind MIMO channel estimation,” IEEE Trans. on Signal
Processing, vol. 54, no. 3, pp. 861–869, 2006.
[1] M Kiessling, J Speidel and Y Chen, “MIMO Channel estimation
in correlated fading environments,”: in proceeding of 58 th IEEE
vehicular technology conference (VTC’03), Orlando,
pp. 1187-91, OCT 2003.
[12] G Xie, X Fang, A Yang and Y Liu, “Channel estimation with
pilot symbol and spatial correlation information,”: in proceeding of
Fig. 4 BER Vs SNR for proposed channel estimation technique
using 4 pilot symbols and 100 Blind symbols using 2 transmitter
IEEE international Symposium on communication and information
and diff. receiver antennas for 4 PSK modulation scheme Technologies (ISCIT’07), Sydney, pp.1003-6, OCT 2007.
[13] S M Kay, Fundamentals of statistical signal processing:
V. CONCLUSION Estimation Theory. Prentice – Hall, Upper saddle River, NJ 07458,
1993.
A new joint semi-blind channel and data estimation [14] G K Krishnan and V U Reddy, “MIMO Communication –
technique is proposed and investigated for Rayleigh faded motivation and a practical realization,” IETE Tech Rev,
MIMO channel matrix H using different receiver antennas Vol.24,No.4,pp.203-13, jul-aug 2007.
combinations. Employing these results, we have [15] T wo and P A Hoeher, “Semi – Blind Channel estimation for
frequency selective MIMO system,” in IST Mobile Summit,
demonstrated that proposed novel technique performs better
Dresden, jun 2005.
than conventional WR and ROML based semi-blind channel [16] T Cui and C Tellambura, “Semiblind Channel estimation and
estimation techniques. Further with increases in receiver data detection for OFDM System with optimal pilot design, “
antennas, BER improves. IEEE Trans. Commun, Vol.55,No.5,pp. 1053-62, may 2007.
[17] T wo, P A Hoeher, A Scherb and K D Kammeyer,”Performance
AKNOWLEDGEMENT analysis of maximum – likelihood semi blind estimation of MIMO
Channel,” in Proceedings of 63 rd IEEE vehicular Technology
We are very thankful to our institute, Chandubhai S. Patel Conference (VCT), Melbourne,pp. 1738-42, may 2006.
Institute of Technology, Charotar University of Science and [18] S Chen, X C Yang, L Chen and L hanzo, “ Blind joint maximum
Technology-Changa for their financial support. likelihood channel estimation and data detection for SIMO System,”
int. J.Auto.Comput., Vol.4, no.1,pp.47-51, jan 2007.

© 2012 ACEEE 38
DOI: 01.IJCOM.3.1.514


[19] K Sabri, M El Badaoui, F Guillet, A Adib and D Aboutajdine, [21] M Abuthinien, S Chen, A Wolfgang, and L Hanzo, “jont
“A Frequency domain based apporch for blind MIMO System maximum likelihood channel estimation and data detection for
identification using second order cyclic statistics,” Elsevier signal MIMO system,”: in proceeding of IEEE international conference
Processing, vol.89, No.1,pp.77-86,jan 2009. on Communication (ICC’07),Glasgow,pp.5354-8,jun 2007.
[20] I M Panahi and venket, “Blind identification of multi-channel [22] M A Khalighi and s Bourennane, “Semiblind single – carrier
system with single input and unknown order,” Elsevier signal MIMO Channel estimation using overlay pilots, “IEEE Trans.
processing, vol.89, No.7,pp.1288-310,jul 2009. Vehicular Tech., Vol.57,no.3,pp.1951-6, May 2008.

© 2012 ACEEE 39
DOI: 01.IJCOM.3.1. 514

Modified Whitening Rotation based Joint Semi-blind Channel and Data Estimation Scheme for Rayleigh Flat Fading MIMO channels

More Related Content

What's hot (19)

Viewers also liked (12)

Similar to Modified Whitening Rotation based Joint Semi-blind Channel and Data Estimation Scheme for Rayleigh Flat Fading MIMO channels (20)

More from IDES Editor (20)

Recently uploaded (20)

Modified Whitening Rotation based Joint Semi-blind Channel and Data Estimation Scheme for Rayleigh Flat Fading MIMO channels