Development of Fuzzy System Based Channel Equalisers
Channel equalisers are used in digital communication receivers to mitigate the effects of inter symbol interference (ISI) and inter user interference in the form of co-channel interference (CCI) and adjacent channel interference (ACI) in the presence of additive white Gaussian noise (AWGN). An equaliser uses a large part of the computations involved in the receiver. Linear equalisers based on adaptive filtering techniques have long been used for this application. Recently, use of nonlinear signal processing techniques like artificial neural networks (ANN) and radial basis functions (RBF) have shown encouraging results in this application. This thesis presents the development of a nonlinear fuzzy system based equaliser for digital communication receivers. The fuzzy equaliser proposed in this thesis provides a parametric implementation of symbolby-symbol maximum a-posteriori probability (MAP) equaliser based on Bayes’s theory. This MAP equaliser is also called Bayesian equaliser. Its decision function uses an estimate of the noise free received vectors, also called channel states or channel centres. The fuzzy equaliser developed here can be implemented with lower computational complexity than the RBF implementation of the MAP equaliser by using scalar channel states instead of channel states. It also provides schemes for performance tradeoff with complexity and schemes for subset centre selection. Simulation studies presented in this thesis suggests that the fuzzy equaliser by using only 10%-20% of the Bayesian equaliser channel states can provide near optimal performance. Subsequently, this fuzzy equaliser is modified for CCI suppression and is termed fuzzy–CCI equaliser. The fuzzy–CCI equaliser provides a performance comparable to the MAP equaliser designed for channels corrupted with CCI. However the structure of this equaliser is similar to the MAP equaliser that treats CCI as AWGN. A decision feedback form of this equaliser which uses a subset of channel states based on the feedback state is derived. Simulation studies presented in this thesis demonstrate that the fuzzy–CCI equaliser can effectively remove CCI without much increase in computational complexity. This equaliser is also successful in removing interference from more than one CCI sources, where as the MAP equalisers treating CCI as AWGN fail. This fuzzy–CCI equaliser can be treated as a fuzzy equaliser with a preprocessor for CCI suppression, and the preprocessor can be removed under high signal to interference ratio condition.