Soft computing approach to adaptive removal of interference signal

A soft computing approach using neural network, fuzzy logic and machine learning to the problem of interference canceling is proposed in the paper. The famous signal-processing structure of adaptive noise canceling is used for the research of interference signal canceling, in which a neuro-fuzzy system is used as the adaptive notch filter. Four T-S fuzzy rules are in the neuro-fuzzy filter. The filter integrates the adaptation capability of neural network and the inference ability of fuzzy logic, so that the signal-processing policy and the meaning of learning are transparent. To explore the excellent nonlinear mapping ability of the neuro-fuzzy adaptive, an appropriate machine-learning algorithm has to be used for the learning purpose, so that the optimal or near-optimal parameter set of the neuro-fuzzy filter can be obtained. The well-known Random Optimization (RO) algorithm and the famous Least Square Estimate (LSE) algorithm are used in hybrid way for the filter. To demonstrate the proposed approach, an exemplar experiment is implemented. A good discussion for the approach is given. The proposed neuro-fuzzy adaptive filter shows great filtering performance.