Channel Equalisation Using Multilayer Perceptron Structure

This paper investigates the application of artificial neural network structures to the problem of nonlinear channel equalization. The difficulties caused by channel distortions such us inter symbol interference (ISI) and nonlinearities can be overcome by equalizers employing neural networks. We present a structure of a nonlinear equalizer with decision feedback using multilayer perceptron (MLP DFE) which is trained by means of the back propagation algorithm. We compare its performance with a conventional decision feedback equaliser (DFE). Simulation results show that the performances of the MLP DFE surpass significantly the classical DFE in term of signal equalised and the steady state mean square error (MSE) attainable. The conventional DFE perform poorly on the nonlinear channel whereas the MLP DFE provide more advanced solution for the nonlinear equalisation problem, the consistency in performance is observed in nonlinear channels.