The Real-Time Dynamic Gesture Recognition Based On Computer Vision Technology And Its Application

In recent years, the real-time gesture recognition based on computer vision technology has been researched widely, the development of this technology brings a more friendly and more convenient human-machine interactive way. It wins the favor of the public gradually, now more and more products which used the gesture recognition technology based on vision have entered into our daily life.In this paper, we study the real-time dynamic gesture recognition technology which is based on computer vision, and used the OpenCV library to design an application based on this technology. The application is used for the presentation control of PowerPoint in the multimedia teaching. The main content of this paper is divided into four parts:gesture detection, gesture tracking, gesture recognition and the design of the application.In the section of gesture detection, firstly, using the DirectShow to achieve the collection work of real-time video image, and then using the motion detection method of the background difference combined with the information of color histogram to completed gesture detection; In the experiment, the brightness values of the color of motion skin region have been made a special adjustment in order to eliminate the influence of the face region which is static.In the section of gesture tracking, the particle tracking algorithm based on the grain of color histogram to achieve the gesture tracking. Because the brightness values of the color of motion skin region have been made a special adjustment in the gesture detection section, so this method can solve the occlusion problem of human face in the motion process well.In the section of gesture recognition, using the template matching method which based on the characteristics of Hu moment invariants. Because the Hu moments of image have the invariance principle of displacement, rotation and scale, so it can be used to descrip the gesture well.In the section of application implementation, we select six kinds of gestures from the gesture-template-library, and then give them different semantics. In the last, we achieve the control operation for power point according to the results of the real-time gesture recognition.