Nonlinear channel equalization for QAM signal constellation usingartificial neural networks

摘要

Application of artificial neural networks (ANN's) to adaptivenchannel equalization in a digital communication system with 4-QAM signalnconstellation is reported in this paper. A novel computationallynefficient single layer functional link ANN (FLANN) is proposed for thisnpurpose. This network has a simple structure in which the nonlinearitynis introduced by functional expansion of the input pattern byntrigonometric polynomials. Because of input pattern enhancement, thenFLANN is capable of forming arbitrarily nonlinear decision boundariesnand can perform complex pattern classification tasks. Consideringnchannel equalization as a nonlinear classification problem, the FLANNnhas been utilized for nonlinear channel equalization. The performance ofnthe FLANN is compared with two other ANN structures [a multilayernperceptron (MLP) and a polynomial perceptron network (PPN)] along with anconventional linear LMS-based equalizer for different linear andnnonlinear channel models. The effect of eigenvalue ratio (EVR) of inputncorrelation matrix on the equalizer performance has been studied. Thencomparison of computational complexity involved for the three ANNnstructures is also provided