Blind deconvolution of linear systems with nonstationary discrete inputs

Abstract

A method is proposed for blind deconvolution (equalization) of communication channels when their input signals are real- or complex-valued multilevel random sequences. The gist of the method is to apply a linear filter (equalizer) to the observed signal and to adjust it until a multilevel sequence is obtained from the output. It is shown that the problem can be solved with only scale/rotation and shift ambiguities. A cost function is proposed so that any minimizer of the function provides a solution to the problem. When the channel is parametric, a procedure is presented for the consistent estimation of the channel parameters. All these results are obtained for nonminimum phase linear systems without assuming the stationarity of the signal.<>