The Research And Implementation Of Hand Gesture Recognition Algorithms Based On Vision

Hand gesture recognition is a important human-computer interaction technology.It provides personal information input and command control for human computer interaction.After analyzing the status quo of gesture recognition,this paper finds out the problems and difficulties in the research of gesture recognition based on vision,from the gesture segmentation,static gesture recognition and dynamic gesture recognition in three aspects of theoretical research,and complete the implementation of gesture recognition system.In view of the existing gesture segmentation methods in complex environment,they are influenced by the influence of light,regardless of skin color and other factors,the segmentation effect is not ideal,the effect of the depth image of light and color is not sensitive to this feature,RGB-D image based on YCr'Cb' elliptic model and K-means is presented in this paper for the method of gesture recognition,position and depth information of skin color area by K-means clustering processing,to eliminate irrelevant color interference;through correction of gesture target using location information,to reduce the impact of changes in illumination.Experimental results show that the proposed method can effectively achieve the segmentation of RGB-D image in complex environment.Aiming at the recognition algorithms that have low recognition rate and robustness problem in target rotation and zoom,a static gesture recognition method based on the tip of the finger points and Hu moment is proposed in this paper,this method using the characteristics of finger,that similar to the shape of the cylinder,by calculating the class width at different positions of the fingers improve the fingertip detection method based on curvature,in order to enhance robustness of the tip of the finger point feature extraction;fuse Hu invariant and finger point features,to improve the global descriptive of the gesture model;the recognition algorithm can automatically add a new gesture,to improve the practicability of the algorithm.Experimental results show that the proposed method can effectively improve the accuracy and robustness of static gesture recognition.In order to solve the problem of low recognition rate and poor real-time performance in the complex environment,a dynamic gesture recognition method based on RGB-D information is proposed.The depth information is introduced into the traditional Kalman filtering algorithm,as one of the parameters to improve the gesture tracking accuracy;through the window detection is added in the tracking process,in order to improve the tracking efficiency;combined with fast dynamic time warping algorithm and highlight key feature points,improve the traditional dynamic time warping algorithm,and applied to gesture trajectory matching,in order to improve the matching efficiency.Experiments show that this method can effectively improve the efficiency of dynamic gesture recognition in complex environment.On the basis of the hand gesture recognition algorithm,the static and dynamic gesture recognition system based on Qt and Open CV is designed and implemented.Static hand gesture recognition system respectively supports RGB image,RGB-D image and gesture image directly through the camera to read the video image of three kinds of data input methods,and complete the gesture segmentation and fingertip contour feature extraction and gesture recognition.