Analysis of BER and MSE performance in nonlinear equalization using modified recurrent network

Nonlinear inter-symbol interference (ISI) leads to significant distortion and performance degradation in wireless digital communication system in presence of additive white Gaussian noise. An adaptive equalizer is used to neutralize the effect of nonlinear ISI and Gaussian noise for better bit-error rate (BER) performance. In this paper, a faster convergent recurrent neural network structure updated by a stable normalized Back-Propagation (RNNNBP) is proposed for nonlinear channel equalization to nullify ISI. The MSE and BER performance of the proposed method are compared with the conventional MLP (feedforward network) and RNN. The nonlinear equalizer presented shows better performance in presence of higher order distorted non-linear models.