Research On Core Technology Of Dynamic Gesture Recognition

With the characteristics of visualized, imagery, vivid and abundant information containment, dynamic gesture recognition which is a part of human-computer interaction (HCT) technology has attracted much attention of researchers and has become one of the hottest research points.In this paper, the basic theory and research actuality of dynamic gesture recognition based on vision are briefly introduced firstly. Then, the recognition algorithm of gesture segmentation algorithm and dynamic gestures trajectory in gesture recognition are discussed. In order to solve the problem that the accuracy in segmentation is low, the hand segmentation algorithm based on the color association movement area is presented. In order to solve the problem that the calculated amount of dynamic gestures trajectory recognition is large, the gestures movement trajectory recognition algorithm based on trajectory subsection characteristics is presented. Finally, we take an experiment about the prescribed dynamic gestures in this paper, the experimental results show that the recognition rate is high.This paper has the following innovation points:In the gesture segmentation respect, firstly, Affinity Propagation Clustering is used to extract key-frame of the video sequence, the problems of low recognition rate and large amount calculation the large amount of calculation which are caused by the improper selection of key frames are solved. Then, the problem about the fusion of the hands and the face existing in the segmentation is well solved by the method of color association movement area that this paper puts.In the recognition of gestures movement trajectory respect, this paper puts forward an algorithm, this algorithm uses the characteristics of segmented trajectory to identify the gesture trajectory. Through by the monotonic property of the trajectory, this method divides the trajectory firstly. Then, the orientative feature of each section curve and the position relation feature about the start position and end position of the trajectory are used to identify the dynamic gestures trajectory. This method not only avoids the curve fitting of movement, but also reduces the time complexity of the algorithm in a large extent.