Applications of neural networks to digital communications - a survey

Neural networks (NNs) are able to give solutions to complex problems in digital communications due to their nonlinear processing, parallel distributed architecture, self-organization, capacity of learning and generalization, and efficient hardware implementation. The paper gives an overview of the applications of NNs to digital communications such as channel identification and equalization, coding and decoding, vector quantization, image processing, nonlinear filtering, spread spectrum applications, etc. The key issue in neural network approaches is to find an appropriate architecture that gives the best results. The paper shows, through several examples, how to choose the neural network structures and how to combine neural network algorithms with other techniques such as adaptive signal processing, fuzzy systems and genetic algorithms. Finally, the paper reviews the mathematical approaches used to understand the learning and convergence behavior of neural network algorithms. (C) 2000 Published by Elsevier Science B.V. All rights reserved.