Blind tree search over non minimum phase channels using cumulants

Maximum likelihood sequence estimation (MLSE) is a possible approach to deconvolution in presence of digitally modulated signals when the channel is afflicted by multipath propagation. The best approach, if optimality is of concern. The optimum MLSE is however severly affected by the channel estimation bias and, this effect becomes much more critical in reduced complexity tree search algorithms. Particularly, channels with non-minimum phase (and this is the common characteristic of real world communication channels) also cause performance degradation if perfectly known. The algorithm we present here is based on higher than second order (HOS) statistics processing. We modify the classical maximum likelihood `metric´ in order to integrate information about the channel and about the likelihood of the current hypotheses. Information is obtained from the received signal (by means of the fourth order cumulants) without the need of any training sequence or signal. The result is a low complexity approach to the blind trellis search