Blind Channel Equalization Using Elman Network

Abstract

Advanced communication techniques sometimes also cause limitations due to destructive influence on digital communication system. Occurrence of noise due to co-channel interference, multipath delays causing channel fading, etc. are the factors that cause the receiver to get corrupted bits. Several equalization techniques have been proposed to remove this effect of Inter-Symbol Interference (ISI) and Multi-User Interference (MUI), but they suffer from some limitations, and their performance is somehow suboptimal. This paper introduces equalization methods and compares their performance and proposes a better solution, which is efficient in removing ISI, moreover saves the bandwidth by eradicating the training sequences. We employ Elman Network, which is a classical recurrent neural network to work as an equalizer. We have showed through simulation that our designed recurrent neural network to work as an equalizer performs better, given that the network is trained properly.