Channel estimation based adaptive equalization/diversity combining for time-varying dispersive channels

We propose a novel reduced dimension channel impulse response (CIR) estimation procedure. Since the overall CIR is a convolution of the transmit and the channel filter, the number of unknown parameters can be reduced also reducing the training sequences. For symbol detection in fast fading, we introduce a new DFE coefficients computation algorithm which incorporates the channel variation during the decision delay into the minimum mean square error criterion, which we refer as to the non-Toeplitz DFE (NT-DFE). Finally, we show the feasibility of a suboptimal receiver which has complexity less than that of recursive least squares type adaptations with performance close to the optimal NT-DFE