The Research Of Hand Gesture Recognition Based On Skin Color And CamShift Algorithm

With the rapid development of computer technology, the Human-Computer interaction ways have become more and more various. And the traditional Human-Computer interaction ways, such as keyboards, mouses, and microphones cannot meet the need any more. Compared with the old interaction ways, the new Human-Computer interaction technology is more "humanized". Hand is the most flexible part on human body, and hand language is considered as the second language of human people. Thus the research on gesture recognition has become a hot for recent years. The gesture recognition relates to the following fields: computer vision, pattern recognition, biometrics and so on.This paper mainly researches the recognition and tracking ways of gestures, and then develops a dynamic gesture recognition system based on this. The users don't need to wear any equipment, and all they need to do is to make the already defined gestures which can be recognized by the system. Because of the application difference of different recognition and tracking ways, there is a difference in the effectiveness of gesture recognition for the system. After an analysis and comparing of several commonly seen gesture recognition and tracking ways, this paper select the most appropriate gesture recognition and tracking way for this system. The result shows that this system has good realtime feature and accuracy. The gesture recognition technology based on monocular vision is usually constituted by the key technologies of gesture modeling, gesture segmentation, feature extraction, and gesture matching. And gesture dividing is the most difficult and important step for the gesture recognition program to recognize the gestures. It's the precondition for computers to do subsequence processing. This system will convert the hand images from RGB space to HSV space, and segment the hands using skin color threshold value. The results prove that skin color segmentation is simple and effective, which is very appropriate for gesture segmentation.The most popular gesture tracking algorithm based on monoculur vision are the particle filtering algorithm and Camshift algorithm. After a brief introduction to these two algorithms, this paper decides to use Camshift algorithm to track gestures. Because Camshift algorithm is semiautomatic, we have to mark the gestures by hand before tracking. Thus this system combines the gesture tracking and recognition technologies, and improves the Camshift algorithm, which solves the semi-automatic problem of Camshift and realize the real-time gesture recognition.