Blind multichannel estimation exploiting the finite symbol alphabet

Unlike the recent works dealing with purely blind channel estimation algorithms that are based on the second-order statistics of the received signal, in this paper we address the exploitation of the finite alphabet of the transmitted symbols to improve the blind estimation performance of the channel. The incorporation of the finite alphabet nature leads the symbols present in the problem to act as a training sequence for the channel estimation. Hence, a blind approach that exploits the symbol alphabet outperforms its purely blind version. We propose to incorporate the prior knowledge of the finite alphabet by combining a purely blind channel estimation criterion with a decision-directed linear MMSE equalization criterion. This combined criterion corresponds to an optimally weighted least-squares approach. Simulation results demonstrate that significant improvement can be obtained by exploiting the finite symbol alphabet.