Vision-Based Gesture Recognition Referring To Body Structure

This article proposed a new method which introduces human body structure parameters into gesture recognition besides traditional vision information. By establishing the subject’s skeleton model, the system could rule out background interference from real-time images quickly, locate the subject’s joints accurately, and track the gestures automatically. This method applies a common monocular webcam instead of expensive depth cameras or complex binocular cameras. It can build a vivid human model and simulate the subject’s gestures well. After grouping nearest classification (GNC), a modified k-nearest neighbor (KNN) classification which has self-learning ability, the relationships between these joints’ parameters are analyzed by the method, and hence the corresponding postures and movements are identified. This method avoids global scanning for each frame and simplifies the classification process, and thus the computation amount decreases sharply. The results show that the method has a good tolerance of dynamic complex background, it can solve the problem of missing targets well.