Visual-gesture recognition with color segmentation and support vector machines
Zhou Hang, Ruan Qiuqi · 2005
We present a two-stage hand gesture recognition system. It includes a novel approach for real-time skin segmentation in video using color space. The approach enables robust segmentation of skin-colored patches despite certain illumination conditions. We give a skin range to broaden the field of skin color to deal with occlusion and effect of background. The performance of our algorithm was approved to the segmentation obtained using static color model. Then we use approved linear SVM learning algorithm to train pixel data set and established a structure decision tree. Finally, the result of recognition system achieved accuracy for testing sequence at 92%.