基于交互式隶属函数调整的模糊神经网络

Fuzzy Neural Networks Based on Adjustment of Interactive Membership Function

对基于模糊神经网络的人脸识别方法进行了研究,提出的模糊神经网络在高维空间(如人脸)上进行模糊规则获取中得到的最终隶属度函数值并不合理,易得到零值的情况,对模糊规则获取过程中的隶属度函数初始值进行人为地调整.在ORL人脸库的实验结果表明,该方法具有良好的性能.

A study is made on the Fuzzy-Neural-Network-based face recognition approaches. Since the final membership function value, if obtained by extracting fuzzy rules on high dimension space （face, for instance）, is not so reasonable and might very well be zero values, human adjustment is made on the initial value of the membership function during the extraction of fuzzy rules. Experimental results on the ORL facial database indicate that the proposed method has excellent performance.