A frequency-domain neural network equalizer for OFDM

OFDM is regarded as a viable solution to combat the impact of frequency selective fading; however, the channel does not have flat amplitude over the entire bandwidth, thus channel equalization is still required at the receiver. Radial basis function (RBF) neural networks have been widely considered for channel equalization, since they offer certain advantages over conventional equalizer structures. In this paper, a novel RBF channel equalizer structure, which performs Bayesian estimation, is proposed for OFDM communication systems. The proposed equalizer structure is shown to outperform existing equalizers; it can therefore be considered as a better practical alternative for OFDM channel equalization.