Multilayer perceptron structures applied to adaptive equalisers for data communications

The authors demonstrate the application of multilayer perceptrons to the problem of equalization in digital communications systems. Following a brief description of the architecture of the multilayer perceptron, the authors investigate its capabilities as a channel equalizer and show that they are superior to those of the linear equalizer. The comparison is effected by simple examples that show graphically why the nonlinear architecture provides better results, and also by an investigation of bit-error rates in a higher dimensional situation. It is concluded that the multilayer perceptron offers advantages over the linear transversal equalizer in a number of areas. Its ability to form decision regions with nonlinear boundaries enables it to equalize both minimum-phase and non-minimum-phase channels without the introduction of any timing delay, a capability that may be of value in nonstationary environments