Fuzzy System Learned Through Fuzzy Clustering and Support Vector Machine for Human Skin Color Segmentation

Abstract

This paper proposes a Fuzzy System learned through Fuzzy Clustering and Support Vector Machine (FS-FCSVM). The FS-FCSVM is a fuzzy system constructed by fuzzy if-then rules with fuzzy singletons in the consequence. The structure of FS-FCSVM is constructed by fuzzy clustering on the input data, which helps to reduce the number of rules. Parameters in FS-FCSVM are learned through a support vector machine (SVM) for the purpose of achieving higher generalization ability. In contrast to nonlinear kernel-based SVM or some other fuzzy systems with a support vector learning mechanism, both the number of parameters/rules in FS-FCSVM and the computation time are much smaller. FS-FCSVM is applied to skin color segmentation. For color information representation, different types of features based on scaled hue and saturation color space are used. Comparisons with a fuzzy neural network, the Gaussian kernel SVM, and mixture of Gaussian classifiers are performed to show the advantage of FS-FCSVM.