Studies On The Gesture Recognition System Based On Computer Vision

The vision based gesture recognition is an important part of new generation Human-Computer Interface. It makes the computer know the meaning of the hand gesture and response for.The theory and the background of the vision based gesture recognition are summarized, and constraint based model is set up. The hand is segmented combined with skin color and a learning algorithm for segment is proposed. After the discussion of Support Vector Machine's promotion, the dynamical tracking and estimation is discussed and according corresponding algorithm. Then, real-time gesture recognition is introduced to. Finally, the thesis is reviewed and a prospect is showed. From the above, Gesture Recognition is proved a high value to the future research.The paper had made the following innovative researches:A skeletal constraint BR Model is proposed. The advantage of the algorithm is the ability of getting rid of minor information, reducing the number of image's feature and saves so much time.A SBB algorithm based on skin color is proposed for hand location. it keeps the fine features of Boosting. Furthermore, occlusion and bend can be overcomed.A skin-color associated flow estimation algorithm is proposed. This method is robust in noise and not sensitive to illumination. It achieved the purpose of getting rid of effect of illumination.A vector coherence mapping tracking algorithm based on skin is proposed. It reduced the affection of the recognition from the blur of gesture images. Meanwhile, a set of blur set are built for constraint.A hierarchy of decision is proposed for the field of SVM of one-to-all. The Hierarchy of SVM proves a lower computation cost and it's suitable for the a little complex real-time condition.