Blind Equalization Using Pseudo-Gaussian-Based Compensatory Neuro-Fuzzy Filters

An important problem in data communications is that of channel equalization, i.e., the removal of interference introduced by linear or nonlinear message corrupting mechanisms, so that the originally transmitted symbols can be recovered correctly at the receiver. In this paper we introduce a Compensation-Based Neuro-Fuzzy Filter (CNFF) based equalizer whose high performance makes it suitable for high-speed channel equalization. The compensatory fuzzy reasoning method is used in adaptive fuzzy operations that can make the fuzzy logic system more adaptive and effective. Besides, the pseudo-gaussian membership function can provide the compensatory neuro-fuzzy filter which owns a higher flexibility and approaches the optimized result more accurately. An on-line learning algorithm, which consists of the structure learning and the parameter learning, is proposed. The structure learning is based on the similarity measure of asymmetry Gaussian membership functions and the parameter learning is based on the supervised gradient decent method. We apply the proposed CNFF to co-channel interference suppression (CCI) and additive with Gaussian noise (AWGN). Computer simulation results show that the bit error rate of the CNFF is close to the optimal equalizer.