Blind Equalization of Constant Modulus Signals Based on Gaussian Process for Classification

Abstract

Blind equalization can be combined with soft-input decoders to greatly improve the performance of communication system. However, most blind equalization algorithms are not designed to provide posterior probability, which is essential for soft-input decoders. In this paper, blind equalization based on Gaussian process for classification (GPC), which could output such information, is applied to optimally detect the constant modulus signals. The scheme is implemented by automatically selecting proper initial training data set and continuously incorporating more appropriate points into training data set with a iteration process. During the iteration process, we utilize all equalizer input symbols that can be assumed to be a certain class label at a high probability as training data to make prediction with GPC model, and give out posterior probability of each input symbol under a specific class label. The proposed blind equalizer has been proved to be able to provide a better performance in both linear channel and nonlinear channel compared with other blind equalizers.