A Real-Time Hand Gesture Recognition And Application Based On Vision

Because of its intuition and convenience,based on computer vision and fused multiple subject theories,hand gesture recognition has become a research hotspot in today’s society.In this paper,the key technology of hand segmentation,feature extraction and classification recognition are studied,and the research results is applied to the video player which based on human-computer interaction control, a natural flexible way of human-computer interaction is implemented.Firstly, in order to adapt to the complex environment, combined with skin information and motion information, a gesture segmentation method is applied to segment the ideal hand gesture.The YCbCr Gauss skin color model is used to extract the skin area from the real-time hand gesture video. Background subtraction method is applied to extract the motion region.After and operation on the skin region and the motion region, a skin motion area is obtained.Morphological methods is adopted to remove the noise of the skin motion region. After the above operations, the exact hand gesture region is obtained.Then, a static hand gesture recognition method based on HOG and SVM is presented.One of the HOG features, named rotary variable is analysised by experimental, and the gestures sample library which contained five kinds of hand gesture is established. Based on the training model, the test samples which have been extracted the HOG feature are recognized. Experimental results show that the different direction of the same hand gesture can be recognized, and the accurate recognition rate is 93.08%.The HOG feature has the characteristics of scale variability, for this problem, combined with the circularity, the 7Hu feature which is not sensitive to the scale is applied as the final gesture features. The SVM classifier is used to obtain a training data model. According to the training model, the real-time video image which has been extracted the 7Hu and circularity feature is recognited, and the accrrate recognition rate is 94.66%.Finally, the real-time hand gesture recognition algorithm based on 7Hu moment,circularity feature and SVM are applied into a simple video player. A video player based on hand gesture is implemented.The achievements of this paper lie in:(1) according to the rotary variable characteristicsof HOG, a static hand gesture recognition algorithm based on HOG and SVM is applied to the recognition of different direction of the same hand gesture;(2) according to the scale invariance of 7Hu features, a real-time hand gesture recognition algorithm based on 7Hu moment, circularity and SVM is proposed;(3) The real-time hand gesture recognition algorithm is applied to a video player, and the function of pause, play, fast forward, fast backward based on gesture recognition are integrated into an video player.