Cluster-based blind nonlinear-channel estimation

Abstract

A cluster-based maximum-likelihood sequence estimator (MLSE) for nonlinear channels was described, which consists of a clustering network and an MLSE implemented by the Viterbi algorithm. The cluster-based MLSE can be used for digital communication through nonlinear finite-length channels because channel mapping estimation is used instead of channel estimation in the conventional MLSE. The clustering network of the cluster-based MLSE, which estimates the channel mapping between the signal input vectors and the noiseless channel outputs, is a supervised network and requires a training sequence. We propose a blind channel mapping estimator to estimate the channel mapping without using the training sequence. The blind channel mapping estimator has a clustering block and a mapping block. The clustering block estimates the channel outputs, which represent the channel mapping, subject to an unknown permutation operation because no training sequence is utilized. That permutation operation is resolved by the mapping block, and therefore, the channel mapping is obtained. Introducing the blind channel mapping estimator into the cluster-based MLSE, a blind cluster-based MLSE for nonlinear channels can be done. Computer simulations of the blind channel mapping estimator and the blind MLSE for nonlinear channels are presented.