Tv Function Remote Control-oriented Visual Gesture Recognition Algorithm

Recent years as the extensive use of Computer, people are more and more familiar with network. But for the urban and rural residents of the country, TV still plays an important role in their leisure time. The question about how to control television conveniently and effectively has become a hotspot. And it has important theoretical value and application prospect by using gesture to control TV. At this background, the main work of this paper is as follows:the preprocessing of gesture image; feature extraction; hand tracking and the algorithm of Support vector machine for gesture recognition and its verification.Firstly, the paper introduces a process, including using the algorithm of Camshift to collect the image with gestures, then using a method to make the gesture from the image, and we can also obtain its feature of the gesture. We should make the image of RGB color space model change into the YCbCr color space model, then use the color component in hand tracking, In order to get a single gesture we may use the segmentation of color image, the good method to get feature extraction is using Hu Moments, it can overcome the uncertainty of rotation and scale better because of Hu Moments feature which is not changed in pace with the image's changement in rotation, translation and scale.Secondly, it discusses the theory of support vector machine and its application. It also has an important point about the multi-classification algorithm, the performance of the algorithms such as One-Against-One and One-Against-Rest used in gesture recognition has been analyzed and verified; at the field of kernel function, it discusses the composition of kernel function and the selection of its parameters. The algorithm of Support vector machine using in gesture recognition has been accomplished. The result of the experiment shows that the recognition rates are different in gesture recognition, as using different kernel functions and different multi-classification methods.Finally, on the basis of all the discussions, we propose a two-tier classification method in gesture recognition, as a combination of hand tracking using Camshift algorithm and gesture recognition with support vector machine. The hand tracking is the first layer classifier, it uses the algorithm of Camshift to make the gesture from the image; Support vector machine is the second layer classifier, its role is to get the accuracy of gesture recognition.