Research And Realization Of Large Screen Gesture Interactive System Based On Kinect

With the development of computer and informatization,human-computer interaction is becoming more and more important in office and in life.The natural user interface has become the trend of human-computer interaction.The interaction between language,action,expression and computer is popular in study at present.Gesture recognition,as a key techonology for human-computer interaction,is flexible,intuitive and simple.It becomes a hot topic in the field of research on interaction of action and computer and more scholars pay attention on gesture recognition.The development of gesture recognition is gradually shifted from early wearing of data gloves to research method based on vision.Currently,most studies on gesture recognition are based on computer vision,the background space information is obtained form camera directly rather than wearing of complex and expensive input devices.As we know,touch screen technology is mature now,but most are applied to small screen operations,such as mobile phones,notebooks and small touch screen.For operation of large screen,interactive method is insufficient.This paper research of dynamic gesture recognition and apply it in large screen to achieve the real-time interaction between human and large screen.Kinect is developed in 2010 by Microsoft,it can quickly and accurately obtain human skeletal node information in a complex background environment.In this paper,Kinect is an input device to obtain the key bone node information of hand,hand wrist,elbow and spine for the experimental feature vector.The main work is as following:Aiming at the research of gesture recognition,the whole process is analyzed,and then a detailed investigation and study of the techniques used in each stage of the identification process are discussed.This paper also describes the powerful fuction of Kinect equipment in acquiring depth information and the status and advantages of Kinect gesture recognition.In this paper,a special judgment method is proposed for the detection of the start and the end points of recognition based on the process of gesture recognition.The starting of dynamic gesture is judged by comparing with a pre-defined static gesture,while the end of a dynamic gesture is judged by calculating the distance of motion of the characteristic node in a continuous multi-frame.In the dynamic gesture recognition,the traditional dynamic time warping(DTW)algorithm is improved for the efficiency and rate of the recognition.Firstly,we use the variance of the feature nodes in the process of motion to carry out the dynamic allocation of weights in order to improve the recognition efficiency.Secondly,the search path of DTW algorithm is analyzed in detail,the search is constrained by the combination of point and line to reduce the calculation and improve the recognition rate.Last,this paper combines the improved DTW algorithm with K-nearest neighbor(KNN)algorithm to identify the gesture to improve the recognition efficiency.Finally,this paper introduces an interactive system.In this system,the partical function of mouse is instead of gesture.You can directly operate the large screen through different gestures and the contents of the entire system running smoothly People real experience to human-computer interaction.