A technique for the blind equalization of digital communications channel (12) relies on the iterative minimization of a cost function known as the J-divergence between a known or assumed probability density function (PDF) of the source (10) and an estimated PDF of a receiver decision (18) derived from the equalizer (16) by minimum-distance mapping. The J-divergence function is defined in terms of the Kullback-Leibler distance between the two PDFs. Minimization is achieved by continually updating both an equalizer tap coefficient vector and the estimated PDF of the decision output signal using a stochastic gradient algorithm applied to the J-divergence cost function.
