Adaptive equalization of a nonlinear recursive channel using Gaussian RBF based network

In this paper, the adaptive equalization of a nonlinear recursive channel has been investigated. The equalizer uses non-linear mapping of Gaussian radial basis function (RBF) to overcome the impairments when a binary antipodal signal is transmitted through the nonlinear recursive channel. The Gaussian RBF based tapped delay line (TDL) neural network, used as channel equalizer, is first given a training sequence to compute the centers in an unsupervised manner using K-means clustering algorithm. The weights of the neural network are then adaptively adjusted using recursive least square (RLS) algorithm. The performance of the trained equalizer has been tested to compute the bit error rate (BER) at various SNR levels and for different number of hidden neurons. Finally the proposed equalizer is compared with perceptron classifier and back propagation neural networks.