A Post-Distortion Based on BP Neural Network Improved by Genetic Algorithm Aided Particle Swarm Optimization for Nonlinear VLC Systems

The post-distortion algorithm is necessary for visible light communication (VLC) systems, which is aiming at reducing the nonlinear interference caused by LED at the transmitter. In this paper, BP neural network is investigated to remove the nonlinearity at the receiver, which is improved by genetic algorithm (GA) aided particle swarm optimization (PSO) aiming at enhancing the training process. As compared with conventional BP, GA-PSO-BP network has been proved to be with faster convergence and high approximate accuracy. Computer simulations are carried out considering the nonlinear characteristics of LED in VLC systems, and it is shown that this post-distortion scheme can obtain better detection property than conventional BP neural networks.