The Research Of A Real-time Gesture Recognition System Based On Computer Vision

In recent years, with the rapid development of computer technology, modern human-computer interactive system has become the focus of research in the field of artificial intelligence. Gesture recognition is an important aspect of human-computer interactive way.Therefore, gestures with the character of simple and efficient have come into our daily life more and more quickly, and the number of products which use the gesture recognition technology based on computer vision continues to increase.The real-time gesture recognition system is divided into three parts: gesture location,gesture recognition, and gesture tracking. It's mainly based on VS 2008 platform, and uses OpenCV library. This paper is aimed at researching the gesture recognition rate, anti occluding effect of gesture tracking, and the implementation of real-time gesture recognition system.In this paper, the main work can be summarized as:First, in the section of gesture recognition, Adaboost cascade classifier which is based on Haar feature is used to train gestures. On the one hand, due to the added new Haar feature template, detection error of the tilting gestures is reduced, on the other hand, due to the use of classified risk in the weight coefficients of Adaboost classifier, the recognition rate of classifier is improved. At the same time, the method of this paper is compared with traditional recognition algorithm, the experiments show that recognition rate is higher in this paper.Secondly, in the section of Gesture tracking, the optimized gesture tracking algorithm based on compression perception is proposed. And this algorithm uses Euclidean between the target rectangular box and the feature rectangular to optimize the weight coefficient of compression sensor, experimental results show that this method can give a better result in the occlusion of gesture tracking.Finally, In the section of system implementation, the trained classifier is applied to gesture recognition, and location results are passed to the section of gesture tracking, once tracking failure appears, the system will relocation, recognition and track. In addition,experimental results show that this system can provide excellent accuracy and real-time performance.