3D Hand Gesture Recognition In RGBD Images

Hand gesture recognition is a research hot spot of computer vision and HCI areas.It has broad application prospects in many fields,such as virtual reality,robot remote welding,intelligent driving,office assistance,entertainment and sign language recognition.With the development of the software and hardware of the computer,depth image captured by depth camera has been widely used in hand gesture recognition because the three-dimensional information of the depth image could provide more useful data.In this paper,the Kinect depth camera is used to capture RGB image and depth image used in hand gesture algorithms.The hand gesture segmentation and recognition of static gestures and dynamic gestures are conducted in this paper.The main research work is as follows:(1)Firstly,the research background and significance of gesture recognition are demonstrated,and the related works of gesture recognition is reviewed.(2)From static gesture and dynamic gesture respectively,the basic theory and process of gesture recognition are demonstrated.Expanding from Gesture segmentation,feature extraction and classification(hand tracking should be added to dynamic hand gesture recognition),The-state-of-the-art algorithms of gesture segmentation,the classification and selection of features and the commonly used classification and recognition algorithms are introduced in detail.(3)A dynamic gesture recognition algorithm based on RGBD images is proposed.To deal with problems of complex backgrounds and illumination changes,the algorithm of gesture segmentation is proposed intergrating skin-color and deep information.The hand region is extracted quickly and accturately.The appearance based features including the palm center,fingertips and the number of fingers are integrated to represent the hand.The decision tree model for hand gesture recognition is constructed.Experimental results show that the proposed method achieves a high recognition accuracy and strong robustness.(4)A dynamic gesture recognition algorithm based on RGBD images is proposed.For the first frame in the input sequence image,hand region is extracted based on static hand gesture segmentation algorithm.The position of the palm center is captured based on the distance transform algotithm in skeleton extraction.Kalman filter is used for hand tracking.Finally,decision tree model is utilized for hand gesture recognition.(5)The main work of the paper is summarized.The main problems and challenges existing in the research of the existing gesture recognition are analyzed,and the research direction of the next step is pointed out.