Gesture Recognition Based On Hidden Markov Model With Binocular Camera

The Human-Computer Interaction (HCI) becomes an important part of human life. Compared with other HCI methods, gesture recognition has the advantages of briefness, abundant types, and distinct movement. Gesture recognition has good application prospect in the field of vehicle-mounted system and electronic games.This paper researched the gesture recognition with binocular camera by Hidden Markov Model (HMM). The issue of gesture recognition with ordinary camera can be well solved by one-dimensional discrete HMM. As the binocular camera has the advantage of providing more gesture information, this paper solved the gesture recognition with multi-dimensional discrete HMM and multi-dimensional continuous HMM.Firstly, this paper researched the gesture recognition with binocular camera by one-dimensional discrete HMM, and the advantages and disadvantages of this method. Secondly, multi-dimensional discrete HMM, which is provided by one-dimensional discrete HMM, is used to recognize gesture with binocular camera. The multi-dimensional discrete HMM has the ability to recognize some complex gestures. Multi-dimensional continuous HMM, which is also provided by one-dimensional discrete HMM, is used to recognize some special gestures. The feature in multi-dimensional continuous HMM is changed to conform the gesture character. Then an approximate computing method is proposed to reduce the computation time. Thirdly, the multi-dimensional threshold model is provided by one-dimensional discrete threshold model, so that they can output the dynamic threshold. Finally, this paper used some typical gesture samples to test the methods, and the result shows the methods have the ability of real-time recognition and good recognition rate.