Research And Implementation On Computer Vision Based Gesture Recognition

With the development of computing technology,people pays more and more attentions on human and computer interaction. While enrich people’s life, computer technology also puts forward higher requirements on human-computer interaction. In recent years, the human-computer interaction and the development of virtual technology, makes the people not satisfied with the current mouse and keyboard type of human-machine interactive mode. Under such background, the way of human-computer interaction more in line with the human habits is more and more active. Over the years, these interact more in line with the human habits have made gratifying achievements. These new type of human-computer interaction, including speech recognition, face recognition, facial expression recognition and gesture recognition, etc., among them, the research of gesture recognition are more focused by people.In this thesis, we conduct a deep discussion on the gesture recognition technology based on computer vision, we proposes a combination of Hu moment and support vector machine(SVM) for gesture recognition method, and an experiment on a gesture digital number of 0-9 is verified.This thesis first discusses some commonly used methods of image preprocessing, and then the gesture image feature extraction and support vector machine(SVM) theory is introduced. Gestures with rotation Angle, and the size of the scale, proportion, and translation and a series of changes, so the study of gesture recognition first priority is to find a way to adapt to the changes of feature vector to describe gestures, Hu moment given in this paper is a kind of image features satisfy these conditions, the characteristics of gesture image has a good adaptability to various transformation. In terms of the choice of gesture recognition classifier, this paper presents a support vector machine(SVM) related theory, the theory evolved on the basis of statistical learning of a new machine learning method, support vector machine(SVM)’s classification effect is good, especially in dealing with only a small sample of training and nonlinear,high dimension of machine learning. In addition this paper also gives a compatible YCb Cr adaptive skin color segmentation and HSV color space model, the model of gestures to color image threshold segmentation is of great significance, effect is good.The main innovations of this thesis is in presents a compatible YCb Cr and adaptive skin color segmentation model of HSV color space, and put forward the combination of Hu moment and SVM in digital gesture recognition with LIBSVM.