On the applicability to correlated sources of a blind channel equalization method robust to order overestimation

Abstract

We consider the blind equalization problem in FIR multichannel models from the second-order statistics of the channel output, with a correlated channel input whose statistics are known to the receiver. The few algorithms that handle colored sources require exact knowledge of the channel order, a drawback since order determination is a difficult issue. Recently, Gazzah et al. (2002) have presented a channel estimator robust to order overestimation. Although their derivation assumed white sources, we show that it can be suitably modified in order to handle colored inputs. The algorithm is still able to blindly compute an FIR pre-equalizer such that the overall response reduces to an FIR transfer function which is known a priori by the receiver, up to a complex phase rotation. Therefore a post-equalizer can be designed in a blind, straightforward manner. The method remains robust to order overmodeling in the correlated source case.