A Study Of Dynamic Gesture Tracking Recognition And Human-Computer Interaction Technology

In this day and age, with the continuous development of computer and artificial intelligence technology, computer has penetrated into every aspect of human life, and become an important part of the human being’s everyday life. Especially in recent years, with the deepening of the research in the fields of face recognition, gesture recognition, and posture control, various results of human-computer interaction technology emerge in endlessly. Among them, the operability and pure naturality in the gesture recognition technology has made the technology be widely used in various fields, such as the body feeling game, robot control, augmented reality, intelligent household, gesture recognition, etc. Gesture recognition technology is changing the way of people’s daily lives gradually.In this paper, gesture segmentation and tracking, feature extraction, dynamic and static hand gestures recognition in the gesture recognition technology are researched respectively, the concrete contents are as follows:1. This paper introduces two different algorithms to accomplish the accurate hand segmentation, that based on the platform of RGB(color) camera and the platform of Kinect depth camera. The algorithm based on the platform of RGB camera, firstly, uses the mouse to choose the skin target, and skin color model is established according to this area. Secondly, it fuses the color information in the image and motion information which is obtained by tracking algorithm. Finally, the hand will be gestured accurately from the complex by using this algorithm. In the means for hand segmentation based on Kinect depth camera, the algorithm fuses skin segmentation algorithm in RGB with the depth histogram of depth map, and segment gestures in depth and color map adaptively.2. In the feature extraction and static gesture recognition stage, two kinds of static gesture recognition algorithms are introduced in this paper: static gesture recognition algorithm based on template matching and multiple features fusion. The algorithm based on template matching, firstly, extracts the gesture window, the gesture direction, perimeter and area of gestures contours and so on, then completes gesture recognitionbased on template matching according to the above features. The algorithm based on multiple features fusion, first of all, extract interested gesture regions in RGB figure and depth map, combined with depth information, then fuse weighted Hu(Hu invariant moments) characteristics of the depth map and HOG(Histogram of Oriented Gradients) of the RGB figure to establish SVM(support vector machine mode) for static gestures recognition, recognition rate of 9 kinds of gestures is as high as 95%.3. In dynamic gesture recognition stage, the algorithm train trajectory samples with HMM(Hidden Markov Model HMM) and recognize mass center orbit of the input gesture.The experimental results show that the algorithm can effectively identify the 6 predefined gesture trajectorys.4. In human-computer interaction stage, gestures are associated with flight visual simulation system through TCP protocol to realize that the gesture controls flight trajectorys of unmanned aerial vehicle(uav) in the virtual system.The algorithm in this paper, realize the gesture recognition of one or both hands combined with SVM(support vector machine mode), based on multiple features fusion. In human-computer interaction, by using of the result of gesture recognition associated with the flight in three dimension visual simulation system through TCP protocol, the algorithm completes the control of flight trajectorys of unmanned aerial vehicle(uav) in real time and accurately through the hands gestures’ transformation.