Equalizer for an IR-wireless LAN using RBF neural networks

The application of a RBF (radial basis function) neural network to an adaptive equalizer at the receiver of a wireless IR-LAN is considered. Fixing the decision threshold and classifying the received binary signals are the main functions of the RBF. The general problem of equalization binary signals, passed through a dispersive channel and corrupted with noise, is briefly described. The characterization of the receiver and the effects of both Gaussian and shot noise over the signals are studied. A possible architecture for the equalizer and a comparison with other classical structures (multilayer perceptron and linear transversal equalizer), as well as simulation results are given. Considerations about the way of reducing computational complexity are proposed