A network of Kalman filters for MAP symbol-by-symbol equalization

A network of Kalman filters (NKF) has been proposed for the equalization of digital communication channels. The NKF-based equalizer relies on the approximation of the a posteriori probability density function (PDF) of a sequence of delayed symbols by a Weighted Gaussian Sum (WGS). The problem is that the number of Gaussian terms in the sum increases dramatically through iterations and thus, is not practical for an on-line processing. We propose a modified NKF-based equalizer which consists in replacing the WGS by a Gaussian density minimizing a certain quadratic criterion. We show that this later is centered on the MAP estimate of the sequence of symbols. The resulting modified NKF-based equalizer is shown to have improved performance and is compared to the optimal MAP symbol-by-symbol equalizer (MAPSE).