JOURNAL OF COMPUTING, VOLUME 2, ISSUE 1, JANUARY 2010, ISSN 2151-9617

HTTPS://SITES.GOOGLE.COM/SITE/JOURNALOFCOMPUTING/
WWW.JOURNALOFCOMPUTING.ORG 145

Simultaneous SNR up gradation, blind channel equalization and phase recovery of

transmitted signals using Autocorrelation Matching method
Sunita Panda1, S. M. Rout2, M. Singh2, U. Mishra2, S. K. Routray2 & S. P. Panigrahi3
1
KIT, Berhampur

2
KIST, Bhubaneswar

3
KEC, Bhubaneswar

Abstract:

The Autocorrelation Matching (AM) method is a second order statistics-based blind MIMO FIR
channel equalization technique designed for wireless communication systems using multiple
receiving antennas.

This paper presents a new cost function and an algorithm for up gradation of SNR in blind
channel equalization as well as phase recovery of QAM signals at the same time. Extensive
simulation results show that SNR upgraded from the existing methods and at the same time
phase of received signal recovered properly.

Key words: AM (Auto correlation), Channel equalization

1. Introduction

Blind MIMO (multiple-input multiple-output) FIR (finite input response) channel equalization
techniques refer to a category of signal processing methods that are designed to recover a
number of input signals distorted by an unknown MlMO FIR channel without using training
signals to identify the channel. The recovery of input signals is in general based on some
statistical information about the input signals, the output signals of the channel, and some
structure information about the channel model.

This paper presents a new cost function and an algorithm for up gradation of SNR in blind
channel equalization as well as phase recovery of QAM signals at the same time.
JOURNAL OF COMPUTING, VOLUME 2, ISSUE 1, JANUARY 2010, ISSN 2151-9617
HTTPS://SITES.GOOGLE.COM/SITE/JOURNALOFCOMPUTING/
WWW.JOURNALOFCOMPUTING.ORG 146

2. Autocorrelation Matching (AM) Method

The Autocorrelation Matching (AM) method is a second order statistics-based blind MIMO FIR
channel equalization technique designed for wireless communication systems using multiple
receiving antennas [I]. It requires (I) the autocorrelation functions of the signals transmitted into
an MlMO FIR channel satisfy a linear shift-independence condition: and (2) the MIMO FIR
channel itself satisfies an FIR invertible condition. With these conditions satisfied, a transmitted
signal is equal to an output signal of a MISO (multiple-input single-output) FIR equalizer applied
to the received signals from the MlMO FIR channel up to a unitary factor and a finite delay, if
and only if the autocorrelation function of the outpul signal equals that of the transmitted signal.

The AM method can be used to design an interference-resist wireless communication system [2],
such as a wireless local area network operating on an unlicensed spectrum where strong
interference may exist. In such system, the signal sent by a transmitter is always filtered using a
special filter and thus has a unique autocorrelation function different from all interfering signals;
it is separated from the interfering signals and equalized using a MISO FIR equalizer applied to
the received signals of multiple receiving antennas at a receiver. The AM method can also be
used to design a distributed wireless MlMO communication system with high spectrum
efficiency [I]. In such system, all transmitters send signals, each filtered using a distinct filter and
thus having a distinct autocorrelation function, to a receiver over the same carrier at the same
time; these signals are separated and recovered using a group of MlSO FIR equalizers at the
receiver, each recovering a transmitted signal.

3. The Problem
Popular blind equalization algorithm, i.e. Constant modulus algorithm (CMA) of Godard for a
self-adaptive equalizer of QAM system [3] makes use of the 4-th order moment of the signal
magnitude in γ, which does not contain the phase information of the signal, so it cannot recover
the phase of a QAM signal. However, the 4-th order moment of the In-phase or the Quadrature
part of a QAM signal does contain the phase information. Channel equalization and phase
recovery can be done simultaneously by minimizing or maximizing a cost function that is based
on the 4-th order moment of the In-phase signal or the Quadrature signal alone, if, the source In-
phase signal and the source Quadrature signal are statistically independent to each other.
JOURNAL OF COMPUTING, VOLUME 2, ISSUE 1, JANUARY 2010, ISSN 2151-9617
HTTPS://SITES.GOOGLE.COM/SITE/JOURNALOFCOMPUTING/
WWW.JOURNALOFCOMPUTING.ORG 147

System model:
The system model used in this paper is shown in figure-1.

Fig.1: The System Model.

: A group of N pre-filters, , ,
: A MXN FIR channel
: A vector of M FIR equalizers for recovery of
: A FIR filter that reverses approx.
: A vector of N information signals, , ,

: A vector of N transmitted signals, , ,

: A vector of M received signals, , ,

: An estimate of .
: An estimate of .
: A vector of M additive noises, , ,

Assumptions:
The following assumptions are made throughout this paper.

AS1: is a wide-sense stationary vector sequence with zero mean and unitary variance. It is
temporally white and spatially uncorrelated, i.e., for .

AS2: is wide-sense stationary vector sequence with zero mean and unknown variance
, , , . It is temporally white and spatially uncorrelated, i.e.,
, , , for .

AS3: and are temporally uncorrelated, i.e., 0 for .

CONDITIONS
JOURNAL OF COMPUTING, VOLUME 2, ISSUE 1, JANUARY 2010, ISSN 2151-9617
HTTPS://SITES.GOOGLE.COM/SITE/JOURNALOFCOMPUTING/
WWW.JOURNALOFCOMPUTING.ORG 148

In order to assure the validity of using the AM method in a mobile environment, the following
conditions must be satisfied.

Channel Condition

In order to assure the existence of a MISO FIR equalizer that can recover with delay
l from a MIMO FIR channel , the channel must satisfy Condition I below.

Condition 1
The (l+I)th column vector in Ħ1 is independent of all column vectors of Ħ2-N and all other

nonzero column vectors of Ħ1, where Ħ2-N = [Ħ2, Ħ3,…., ĦN]

In order to equalize a signal from a MIMO FIR channel using a blind equalization
method, we need a stronger condition on the channel than Condition 1. It is stated below.

Condition 2

th
The (l+1) column vector in Ħ1 is independent of all column vectors of Ħ2-N ; all nonzero column

vectors of Ħ1 are linear independent.

Note that the above conditions are weaker than the necessary conditions often cited in the
literature of blind MIMO FIR channel equalization. The above conditions imply that one or more
signals can be recovered from a MIMO FIR channel even though the entire MIMO FIR channel
cannot be reversed using a MIMO FIR equalizer.

Signal Condition

In order to assure that a transmitted signal can he recovered by an output signal of a MISO FIR
equalizer by matching the autocorrelation functions of them, the transmitted signals need to
satisfy Condition 3 below.

Condition 3
JOURNAL OF COMPUTING, VOLUME 2, ISSUE 1, JANUARY 2010, ISSN 2151-9617
HTTPS://SITES.GOOGLE.COM/SITE/JOURNALOFCOMPUTING/
WWW.JOURNALOFCOMPUTING.ORG 149

The set of shifted autocorrelation functions of these signals, 1 | 1,2, ,

*
0, 1, 2, is linearly independent for |k|>Le, where rn(k)=E{sn(t)sn (t-k)}. This is called a

linear shift-independence condition.

4. COST FUNCTION AND ALGORITHM

Consider a base-band equivalent system that combines channel and equalizer:

∑ (1)

Where, is the impulse response of the combined system and { } is the source signal
sequence. For QAM signaling, both and y(n) are complex sequences:
and where the subscript I represents In-phase signal and the
subscript Q represents the Quadrature signal, and also is a complex sequence.

Based on the statistic properties of the source signal, the 4-th order moment of can be
calculated as follows:

3 ∑ 3 ∑ (2)

Depending on whether the value of ( 3 ) is greater or less than zero, the will
take its maximal or minimal value .

An equalizer that can achieve the above result is called, a perfect equalizer for the channel, and
such an equalizer exists if certain conditions are satisfied [4][5]. The channel-equalizer combined
system with the impulse response given in above equation already removes all the ISI and
recovers the phase up to π/2 ambiguity. Using differential coded QAM signals, the phase
ambiguity can be solved which are insensitive to a phase error of π/2.

Now we will take the new cost function

(3)

Subject to:

∑ (4)
JOURNAL OF COMPUTING, VOLUME 2, ISSUE 1, JANUARY 2010, ISSN 2151-9617
HTTPS://SITES.GOOGLE.COM/SITE/JOURNALOFCOMPUTING/
WWW.JOURNALOFCOMPUTING.ORG 150

Fig2: BER Vs SNR using proposed cost function

Fig3: Performance of AR function Algorithm

To use the above cost function, we need to find methods to keep ∑ being constant.
We can use Lagrange Multiplier method and corresponding cost function is

∑ (5)

In order to determine the value of , as in practical communications systems, (m4-3m2)≤0 and let

c=1; then the above equation

3 ∑ 3 ∑ ∑ 1 (6)

when λ=-2m4;the above function can reach a minimal with∑ 1; Corresponding cost

function is

2 (7)
JOURNAL OF COMPUTING, VOLUME 2, ISSUE 1, JANUARY 2010, ISSN 2151-9617
HTTPS://SITES.GOOGLE.COM/SITE/JOURNALOFCOMPUTING/
WWW.JOURNALOFCOMPUTING.ORG 151
2
Last term in above equation is a constant and can be replaced by (m4/m2) ; hence the cost

function

(8)

Instantaneous gradient of above equation gives required equalizer updating.

5. SIMULATION RESULTS
We carried out a numerical simulation that tests the algorithm using the equalizer update formula
of Eq. (8). Equalizer is of T/2-spaced FIR structure. The source signals are uniformly distributed
random sequences (16QAM modulated) and the channel is:

0.5294 0.3219 0.6241 (9)

There is a white noise signal applied to the channel in simulation and the SNR is 35dB.

In each simulation, 100 bursts are tested. In each burst test, two uncorrelated white random
sequences on [-I, I] are generated as two information signals, each of which has 10000 symbols
and 10 of them serve as train symbols for the purpose of identifying the delay and polarization
after the signal is recovered.

Fig4:.Signal Constellations: (a) Received signal, (b) Equalized with our algorithm
The original received signals are "eye-closed" (see the constellation shown in Fig.4 (a)) due to
the ISI introduced from the channel distortion. The proposed algorithm can finally recover the
phase error as in fig. 4(b).

Result Analysis:
1. Performance up gradation in terms of improved SNR as may be seen from Fig2 &3.
2. Phase of the received signal is completely recovered as in fig4.
JOURNAL OF COMPUTING, VOLUME 2, ISSUE 1, JANUARY 2010, ISSN 2151-9617
HTTPS://SITES.GOOGLE.COM/SITE/JOURNALOFCOMPUTING/
WWW.JOURNALOFCOMPUTING.ORG 152

3. There is a limitation on this cost function, i.e., the source In-phase signal the source
Quadrature signal must be statistically independent to each other. This limitation can be
overcome by using suitable scrambling techniques at the transmitter.
6. Conclusion
A new cost function and one algorithm developed in this paper for channel equalization.
The proposed system outperforms the existing systems as seen from the results.
REFERENCES

[I] Hui Luo, Ruey-Wen Liu, Xieting Lin, Xin Li, "The Autocorrelation Matching Method for
Distributed MIMO Communications over Unknown FIR Channels", in Prm. of ICASSP, Salt
Lake City, UT, USA, 2001.

[2] Hui Luo, Ruey-Wen Liu, "Apply Autocorrelation Matching method to Outdoor Wireless
LAN on Co-Channel Interference Suppression and Channel Equalization", in Pmc. .of WCNC
2002, Orlando, FL, USA, March, 2002.

[3] D.N. Godard, "Self-recovering equalization and carrier tracking in two dimensional data
communication systems," IEEE Trans on Communications, Vol.Com28, No.11, pp.1867-1875,
Nov 1980.

[4] Y.Li and Z.Ding,"Convergence Analysis of Finite Length Blind Adaptive Equalizers," IEEE
Trans. on Signal Processing, vol.43, No.9, Sept’95.

[5] Y.Li and Z.Ding, "Global convergence of fractionally spaced Godard(CMA) adaptive
equalizers," IEEE Trans. on Signal Processing, Vol.44, No.4, pp. 814-826, April 1996.