A network of adaptive Kalman filters for data channel equalization

The aim of this paper is to revisit the Kalman filtering- based approach of channel equalization. Indeed, Kalman equalizers have already been proposed in the literature as an alternative to more classical structures. However, these Kalman solutions are based on the assumption of Gaussian signals that is not valid in the context of data channel equalization. From an approximation of the density functions of the data signals by a weighted sum of Gaussian probability density functions, we here propose a new structure of an equalizer that is based on a network of Kalman filters operating in parallel. An adaptive version of this network is investigated. It includes the on-line estimation of the channel and of the noise variance.