Data-dependent filters with fuzzy-neural network
Akira Taguchi, Hironori Takashima · 2002
This paper presents a design method of data-dependent filters by using fuzzy inference for the purpose of restoring signals degraded by additive noise. Since the antecedents of fuzzy inference can be composed of many local characteristics, it is possible for the proposed filter to adjust its weights to adapt to local data in input signal. The proposed filter achieve maximum noise reduction in uniform areas and preserve details of input signals as well. Furthermore, the proposed filter can be constructed by fuzzy neural networks, and so the tuning of this results in backpropagation algorithm.