Research on gesture recognition based on support vector machine

Abstract

In response to the problems of low recognition accuracy and large computation of traditional neural network algorithms, a gesture recognition detection model is designed by using human skin color features and SVM model with gesture classification recognition as the target. The method adopts bilateral filtering and other graphical processing to smooth the edges of the palm, detect and binarize the skin color of the region, and filter the binarized image to smooth the edges. The skin color space is transferred from the RCB space to the YUV space under which the gesture region is separated from the background, and morphological processing techniques are introduced in terms of gesture integrity to effectively fill the black hole region and remove the white dot region in the gesture picture, and directly edge the gesture picture. After obtaining the standardized image, the feature values are obtained. The experimental results show that the method improves the accuracy of gesture recognition compared with traditional algorithms.