Semi-blind nonstationary channel estimator based on parallel LMS filtering

A network of extended Kalman filters (NEKF) was proposed in [6] for joint symbol/channel MMSE estimation. In this paper, we propose to transform the NEKF-based equalizer into a Network of LMS Filters (NLMSF) estimator by approximating the prediction error covariance matrix of each branch of the network by a diagonal matrix, so that reducing the corresponding complexity. The so obtained symbol/channel estimator is similar to the one presented by Iltis in [2]. Simulations illustrate the good tracking capacity of the blind NLMSF estimator in a non-stationary environment. A study of the stability of such blind algorithms towards the channel coefficients initialization is an imminent perspective of this work.