Blind equalization and estimation of channel using artificial neural networks

In this work a new technique is presented for blind channel equalization. Most of the existing techniques perform channel estimation in first phase and equalization in second phase. The algorithm presented here provides not only the direct blind equalization of the channel outputs but also provides the estimated channel in parallel. This technique utilizes three layered artificial neural networks (ANN) model accompanied with learning algorithm for updating of the weights. This learning algorithm utilizes the Euclidean distance error as well as the statistics to be maintained a! three different layers of ANN. The weights between first two layers provide an equalization matrix and output of second layer gives estimate of source symbols. The weights between second and third layer provide channel estimate.