Blind estimation of multiple digital signals transmitted over multipath channels

Abstract

A challenging signal processing problem is the blind joint space-time equalization of multiple digital signals transmitted over multipath channels. This problem is an abstraction of a PCS wireless communication scenario. A number of authors have recently proposed algorithms for the blind separation and equalization of multiple co-channel digital signals transmitted through wireless environments with long delay spreads. These algorithms compute the coefficients of a joint space-time equalizer in which the outputs of multiple antennas are oversampled and linearly combined to produce estimates of the transmitted finite alphabet symbols. We present improvements in three directions, namely (1) a more efficient way to do the inherent subspace intersections, (2) a way to solve the problem for channels with largely differing and ill-defined delay spreads, and (3) algorithm-independent lower bounds for the number of antennas and amount of oversampling to provide sufficient resolution in the case of bandlimited signals. The improved algorithm is tested on experimental data.