IMPROVING DESIGN FEEDBACK EQUALIZER PERFORMANCE USING NEURAL NETWORKS

Novel equaliser structures combining traditional transversal equalisers and neural computation have been introduced for adaptive discrete-signal detection. Extensive simulations using a two-path channel model and 16QAM modulation have been run to investigate the performance characteristics of these neural equalisers. The results have shown that they adapt very well to changing channel conditions, including both linear multipath and nonlinear distortions. The new structures are superior when compared to the traditional equalisers with equal computational complexity, especially in difficult channels.