Performance comparison of blind and non-blind channel equalizers using artificial neural networks

In digital communication systems, multipath propagation induces Inter Symbol Interference (ISI). To reduce the effect of ISI different channel equalization algorithms are used. Complex equalization algorithms allow for achieving the best performance but they do not meet the requirements for implementation of real-time detection at low complexity, thus limiting their application. In this paper, we present different blind and non-blind equalization structures based on Artificial Neural Networks (ANNs) and, also, we analyze their complexity versus performance. Since the activation function at the output layer depends on the cost function with respect to the input, in the present work we use mean squared error as loss function for the output layer. The simulated network is based on multilayer feedforward perceptron ANN, which is trained by utilizing the error back-propagation algorithm. The weights of the network are updated in accordance with training of the network to improve the convergence speed. Simulation results demonstrate that the implementation of equalizers using ANN provides an upper hand over the performance and computational complexity with respect to conventional methods.