Multilayer perceptron based decision feedback equalisers for channels with intersymbol interference

The paper describes the application of multilayer perceptrons to the problem of adaptive channel equalisation in digital communications systems. In particular, the use of decision feedback structures is investigated for time-invariant and time-variant bandlimited channels. Simulation results show that the equaliser based on the multilayer perceptron provides better bit error rate performance compared with the conventional decision feedback equaliser if the equaliser length corresponds to the time spread of the channel. Increasing the equaliser length leads to equivalent bit error rates for the decision feedback and the multilayer perceptron based equaliser. Because of the more complex structure of the multilayer perceptron its adaptive behaviour for time-variant channels is inferior compared with the decision feedback equaliser. Nonlinearities on the transmission channel can be equalised much better by the multilayer perceptron based equaliser.