Reduced-decision feedback FLANN nonlinear channel equaliser for digital communication systems

A reduced-decision feedback functional link artificial neural network (RDF-FLANN) structure for the design of a nonlinear channel equaliser in digital communication systems is proposed. When functional expansion utilities are used, the RDF-FLANN does not need the hidden layers that exist in most MLP-based equalisers. So the RDF-FLANN exhibits a much simpler structure than the traditional DFFLANN and thus requires less computation during the training mode. The use of direct decision feedback can greatly improve the performance of FLANN structures. Comparisons of the mean squared error (MSE), the average transmission symbol error rate (SER) and the eye patterns among RDF-FLANN, FLANN and MLP are presented. Simulation results have demonstrated that RDF-FLANN presents the best performance among the three structures.