An eigenanalysis-based method for blind channel identification and equalisation

Abstract

Subspace (SS) methods are effective approaches for blind channel identification, for they achieve a good performance with a relatively short data lengths and work well at low signal-to-noise ratio (SNR). However, they require accurate channel order estimation, which is difficult in a noisy environment. Although linear prediction (LP) methods can handle the problem of channel order overestimation, their performance degrades dramatically when SNR is low. In this letter, we proposed a blind channel identification and equalisation algorithm, based on the eigenanalysis of shifted correlation matrices of the received data and their associated properties. The algorithm is robust to channel order overestimation and not sensitive to noise as well. Furthermore, the algorithm does not require the computation of the correlation matrix pseudo-inverse, as with linear prediction algorithms, nor are the whole noise or signal eigen vectors necessary to achieve identification as with the subspace algorithm, so it is computationally efficient. Copyright © 2005 AEIT.