论文部分内容阅读
In this paper, a fuzzy neural model is proposed for face recognition. Each rule in the proposed fuzzy neural model is used to estimate one cluster of pattern distribution in a form, which is different from the classical possibility density function. Through self-adaptive learning and fuzzy inference, a confidence value will be assigned to a given pattern to denote the possibility of this patterns belongingness to some certain class/subject. The architecture of the whole system takes structure of one-class-in-one-network (OCON), which has many advantages such as easy convergence, suitable for distribution application, quick retrieving, and incremental training. Novel methods are used to determine the number of fuzzy rules and initialize fuzzy subsets. The proposed approach has characteristics of quick learning/recognition speed, high recognition accuracy and robustness. The proposed approach can even recognize very low-resolution face images (e.g., 7x6) well that human cannot when the number of subjects is not very large. Experiments on ORL demonstrate the effectiveness of the proposed approach and an average error rate of 3.95% is obtained.
Each rule in the proposed fuzzy neural model is used to estimate one cluster of pattern distribution in a form, which is different from the classical possibility density function. Through self-adaptive learning and fuzzy inference, a confidence value will be assigned to a given pattern to denote the possibility of this patterns belongingness to some certain class / subject. The architecture of the whole system takes structure of one-class-in-one-network (OCON) , which has many advantages such as easy convergence, suitable for distribution application, quick retrieving, and incremental training. Novel methods are used to determine the number of fuzzy rules and initialize fuzzy subsets. The proposed approach has characteristics of quick learning / recognition speed, high recognition accuracy and robustness. The proposed approach can even recognize very low-resolution face images (eg, 7x6) well that human can not when the n umber of subjects is not very large. Experiments on ORL demonstrates the effectiveness of the proposed approach and an average error rate of 3.95% is obtained.