A Research On Hand Gesture Recognition Algorithm And Design Of Interactive Prototype System For Intelligent

Gesture is a natural and direct mode of communication, with a trend of human-computer interaction to be human-oriented, researches in vision-based gesture recognition become hot subject. This thesis designs and realizes an interactive system of intelligent classroom based on hand gesture recognition with an emphasis of a vision-based hand gestures recognition algorithm from point of image pre-processing, hand segmentation, feature extraction and recognition of hand gestures image.Being obtained by computer camera, the images are filtered by median filter to eliminate the sharp noise and gradient sharpening to enhance edge.In the hand gesture segmentation part, a rough area of hand gesture is firstly obtained by skin color detection and threshold technology; foreground is then divided by using motion information; finally, an exact area of hand gesture is obtained by making use of operation "AND" on the two parts. With the purpose of better segmentation, an operation of morphological filtering is applied.In the feature extraction, an algorithm based on two-dimensional geometric feature is presented with a real-time and robust requirement. The feature is invariant to general transformation including translation, scale and rotation transformation.In the feature recognition part, with the consideration of less defined gestures and low search depth, classification decision tree is applied. The classification decision tree is build to classify the hand gestures according to the threshold of feature which is obtained by training the sample images.The experimental results show that the algorithm based on two-dimensional geometric features and classification decision tree is efficient and the recognition rate is up to 92%.