Blind Identification of Series-Cascade Nonlinear Channels

Identification of nonlinear channels represented with series-cascade models of Wiener and Hammerstein type is considered in this paper. The approach proposed herein is based on a bilinear decomposition of the received signal measurements matrix. Thanks to an input preceding inducing redundancy, one of the factors involved in the bilinear decomposition has a Vandermonde structure. Uniqueness conditions of the bilinear decomposition are derived by assuming that the input signal belongs to a finite alphabet. Then, a new blind identification method is proposed using an alternating least squares (ALS) approach, and some simulations results are presented to illustrate the behavior of the proposed method