Browse Prior Art Database

EM-based Semi-Blind Estimation of Time-Varying Channels

IP.com Disclosure Number: IPCOM000010711D
Original Publication Date: 2003-Jan-13
Included in the Prior Art Database: 2003-Jan-13
Document File: 5 page(s) / 108K

Publishing Venue

Motorola

Related People

Laurent Mazet,: AUTHOR [+4]

Abstract

OFDM systems traditionally perform channel coeffi- cients estimation relying on known training sequences. However in wireless systems, performance and mobility can be further enhanced either by operating semi-blind channel estimation refinement or variation tracking between reference symbols. In this paper, we propose an EM-based channel tracking method using a simple order 1 AR modelling of the channel variations. The proposed approach is applicable to both single carrier and OFDM systems. Simulation results presented in the context of 5 GHz WLANs show that this new algorithm allows to reach the performance of a receiver having a perfect knowledge of the channel coefficients even in the mobile environment (3 m/s).

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 28% of the total text.

EM-based Semi-Blind Estimation of Time-Varying Channels

Laurent Mazet, Véronique Buzenac-Settineri, Marc de Courville, Pierre Duhamel

Abstract

OFDM systems traditionally perform channel coeffi-

cients estimation relying on known training sequences.

However in wireless systems, performance and mobility

can be further enhanced either by operating semi-blind

channel estimation refinement or variation tracking between

reference symbols. In this paper, we propose an

EM-based channel tracking method using a simple order 1

AR modelling of the channel variations. The proposed approach

is applicable to both single carrier and OFDM systems.

Simulation results presented in the context of 5 GHz

WLANs show that this new algorithm allows to reach the

performance of a receiver having a perfect knowledge of

the channel coefficients even in the mobile environment

(3 m/s).

1 Introduction

One of the advantages of OFDM systems is their ability

to provide a low arithmetical complexity equalization

scheme for frequency selective channels. Thanks to the

introduction of cyclic redundancy between consecutive

transmitted blocks, the equalization is reduced in the frequency

domain after demodulation by the FFT to scalar

multiplications, one per each carrier.

Channel coefficients estimation is usually performed using

known training sequences periodically transmitted (e.g.

at the start of each frame), implicitly assuming that the

channel does not vary between two training sequences.

Thus in order to enhance the mobility of wireless systems

and cope with the Doppler effects, reference sequences

have to be repeated more often resulting in a significant

loss in useful bitrate. An alternative is to track the channel

variations by refining the channel coefficients blindly

using the training sequences as initializations for the estimator.

A common way to design semi-blind estimation algorithms

is to use the Expectation-Maximization (EM) algorithm

[10], which is a two-step iterative procedure maximizing

the likelihood function.

Several authors have already proposed EM-based semiblind

channel estimation methods, either in the OFDM

context [7], [11], [9] or not [3]. In particular, in [9], we

proposed a new approach specifically designed for OFDM

systems and taking into consideration the fact that the

guard interval duration is larger than the channel memory.

In this paper, we propose an alternative to the method of

[9] (which can eventually be used jointly) well-suited for

time-varying channels. Namely, it involves a simple AR

modelization of the frequency domain channel coefficients

variations that is taken into account in the EM cost function.

Simulations run over different mobile channels in

the context of a 5 GHz WLAN (Wireless Local Area Network)

show that this simple model allows to reach the performance

of a receiver with a perfect channel knowledge.

Note that though this paper is presented in an OFDM context,

the proposed method is not restricted to OFDM systems

but is also applicable to any single carrier system in a

flat fading channel.

The paper is organized as follows: section 2 introduces

the OFDM system studied and settles the various notations.

Section 3 recalls the EM algorithm and its application

semi-blind channel estimation. Section 4 presents the

new EM-based semi-blind channel estimation method. Finally,

section 5 illustrates the performance of the proposed

algorithm through simulation results.

2 Definitions and notations

Consider a classical cyclic prefix based OFDM transmitter

depicted on figure 1, using N sub-carriers among

which M carry information symbols, the remaining N _M

side ones being fed by zeroes.

...

Mapping

b1(1)

b2(1)

bP(1)

x(1)

...

Mapping

b2(2)

bP(2)

b1(2)

x(M)

...

Mapping

b1(M)

b2(M)

bP(M)

x(2)

...

0

0

...

0

0

I

F

F

T

P

/

S

Channel

...

...

...

...

Gaussian noise

Parallel to Serial

Fourier Transform

Inverse Fast Add

null sub-carriers

Figure 1: OFDM transmitter

Notations are illustrated on figure 1: for 1 _ m _ M, P

c

2003 Motorola, Inc.

bits (bl(m))1_l_P are mapped onto constellation complex

symbol x(m) (e.g. aQAM2P). The resulting frequency domain

vector X = [x(1) _ _ _ x(M)]T is then modulated back

to time domain by IFFT processing and each of its components

is sent sequentially through the channel after a cyclic

extension of length L.

OFDM systems are designed so that the guard interval

length is greater that the channel duration, thus in the frequency

domain the channel can be seen as N parallel flat

fading channels [6].

Thus, for each carrier m we have:

y(m) = H(m)x(m)+n(m)

where H(m) is the frequency domain channel coefficient

on carrier m, and n(m) a gaussian noise. In the sequel, we

will skip the index m and consider a single carrier in a flat

Rayleigh fading:

y = Hx+n

3 EM-based semi-blind channel estimation

of flat fading channels

The EM algorithm [10] is a well-known parametric estimation

method aiming at deriving the parameters q of a

system from the observations of its output y without kn...