Multistep linear predictors-based blind equalization of FIR/IIR single-input multiple-output channels with common zeros (Q2723859)

The aim of the paper is to investigate direct blind MMSE (minimum mean-square error) equalization of SIMO (single-input multiple-output) channels using only second-order statistics of the data and the multi-step linear prediction approach. The author proposes a novel algorithm that is applicable to IIR (infinite impulse response) channels and to SIMO systems having common zeros among the various channels, as long as the common zeros are minimum phase. In the case of non-minimum-phase zeros, an all-pass filtered version of the original input is recovered. Three computer simulation examples involving a 4-QAM signal illustrate the proposed approach, compare it with that of the linear prediction approach and of the subspace approach, and show that the multi-step linear predictors-based MMSE equalizer outperforms the one-step linear predictor-based MMSE equalizer by a wide margin.