Rule Extraction for Dynamic Hand Gesture Recognition using a Modified FMM Neural Network

Abstract

In this paper, we present a rule extraction method using a modified fuzzy min-max neural network for dynamic hand gesture recognition. We introduce a feature relevance measure for the pattern classification based on FMM neural networks. During the learning process, the feature distribution information is utilized to compensate the hyperbox distortion which may be caused by eliminating the overlapping area of hyperboxes in the contraction process. We define a feature saliency measure that represents a degree of relevance of a feature in a classification problem. From the measure, we can classify excitatory features and the inhibitory features which can be used for the rule generation process.