Based Dynamic Gesture Recognition Of The Optical Flow Calculation And The Dtw Algorithm

Gesture recognition includes many science domains, such as: Image Processing, Pattern Classification, Computer Vision and Artificial Intelligence, etc. The main purpose of this study is to apply gestures into Human Computer Interface (HCI), and realize a more natural, harmonious Human Computer Interface. Gesture recognition can be widely applied to virtual reality, three dimensional design, telepresence, visualization, medical research, and the smart surveillance, and have very good social and economic benefits.On the basis of analyze various features in video series, optical-flow features and the main current optical-flow algorithms are introduced in this thesis. And three important optical-flow algorithms: Horn-Schunk, Lucas-Kanade, and Gaussian pyramid optical-flow algorithm are studied and analyzed. Then optical-flow features of gesture video series are extracted and quantizd. And based on these experimental results, the common spatiotemporal template and the index sequences warehouse are built. At last, optimization DTW algorithm is used to realize dynamic hand gesture recognition.In the study of optical-flow algorithm, the software Matlab is used to realize the simulation experiment. As the result of the analyzation of three different algorithms and its experimental results, the Gaussian pyramid optical-flow algorithm can get an more precisely optical-flow field. The recognition system is constructed under the environment of Visual Studio. Net 2008 and OpenCV 1.0, and video series' pretreatment, Gaussian pyramid optical-flow algorithm, DTW algorithm are realized by programming.Experimental results show that, the dynamic hand gesture recognition system can recognize four special gestures easily, and the recognition rate on four gestures is 94.5%.