The Application of Modified Generalized Dynamic Fuzzy Neural Network (M-GDFNN) in Adaptive Noise Cancellation

摘要

This paper proposes a modified generalized dynamic fuzzy neural network (M-GDFNN) for the application of Adaptive Noise Cancellation (ANC). As system error cannot be evaluated on-line in ANC, ε-Completeness of fuzzy rules is selected as the only criteria of fuzzy rules generation in this novel algorithm. To improve system performance, a simple Self-organizing Mapping (SOM) technique is introduced to adjust the center clustering. Additionally, Eigenvalue Decomposition (ED) method is applied to prune unimportant rules. In the last part of this paper, simulation studies illustrate that proposed M-GDFNN is able to achieve better results compared with ANFIS and standard GDFNN in noise cancellation thus prove the suitability of proposed M-GDFNN in the field of ANC.