Artificial Neural Network based interference mitigation through nonlinear channel equalization

This paper suggests the usage of Minimal Resource Allocation Network (MRAN) algorithm based an artificial neural network for mitigating the inter symbol interference through nonlinear channel equalization. The study results provide the application of non-linear channel equalization scheme for data communications, using the structure of minimal radial basis function neural network. The MRAN challenge technique uses online learning with growing and prunes the capability of radial basis function network´s hidden neurons, establishing a parsimonious network structure. Analyzed to existing linear methods, namely Least Mean Square and Recursive Least Square, the proposed methods do not have to calculate the order of the channel first and set the model parameters. The MATLAB results showing the superior performance of the MRAN algorithm for two different non-linear channel equalization problems, with a linear, non-minimum phase and mixed phase problems are presented.