Research on Variable Step-Size Blind Equalization Algorithm Based on Normalized RBF Neural Network in Underwater Acoustic Communication

摘要

In this paper, based on constant modulus algorithm (CMA), variable step-size blind equalization algorithm based on normalized radial basis function (RBF) neural network is proposed, considering blind equalization can equalize nonlinear characteristic of underwater acoustic channel without training sequence and RBF neural network is a nonlinear system with excellent approximation characteristic and performance of equalizing nonlinear channel. The algorithm is emulated in SIMULINK and verified its feasibility and performance using data of lake testing. Simulation and testing results show that variable step-size blind equalization algorithm based on normalized RBF neural network is better than classical BP algorithm and RBF algorithm in convergence rate and equalization performance.