Approximation and Adaptive Equalization of RBF Neural Network in Nonlinear System

In wireless communication system, the interference caused by time-varying and dynamic changes of nonlinear channel may lead to signal distortion and spectrum fading. A genetic algorithm based on decision feedback for adaptive neural network equalizer structure and error back propagation is proposed. The RBF neural network is used in the feedforward part of the equalizer to approximate the nonlinear system channel, the feedback part uses decision feedback based on the recursive least squares algorithm, which is used to quickly capture in the training phase and eliminate the residual inter-symbol interference caused by the feedforward equalizer. The mean square error convergence characteristic and the bit error rate fluctuation curve of the phase modulation signal after adaptive equalization are simulated, and the validity of the equalizer structure and algorithm is verified. The decision feedback RBF neural network adaptive equalizer can effectively eliminate the channel interference of the nonlinear system, and improve the performance of the equalizer in the steady state mean square error and bit error rate.