Design And Implementation Of Office Health Analysis System Based On Kinect

With the development of the fields of Computer Vision, Sensor Technology, Pattern Recognition and Artificial Intelligence for nearly half a century, the center of Human-computer Interaction(HCI) has turned from computer to human. The human-centered interactive mode is obviously based on testing and recognition of the information of human such as action and gesture. HCI can be used in every aspect of our daily lives. As the working intensity increases observably in office nowadays, cervical and lumbar spondylosises become common diseases of office workers. Office heath and ergonomics attract widespread concern among the society. An office health analysis system is designed and implemented in this thesis based on Kinect depth camera, which is a product prototype, aiming at the analysis of office health status and offering recommendations.In this thesis, the development platform based on Kinect is built by connecting the Kinect for Windows depth sensor, Kinect for Windows SDK and Visual Studio 2010 Pro. Since sitting position is the most common position during working in office, the first mission of this system is to test sitting positions by obtaining and computing the joints data of the Kinect platform. Sitting position is determined by cervical angle and lumbar angle. The time periods and frequencies of unhealthy sitting positions and leaving are counted to offer suggestions of office health.Besides sitting positions, behaviors in the office include common actions such as calling, drinking and typing. The frequencies of these actions also reflect the habit of work. Thus gesture recognitions of calling, drinking and typing are implemented based on Kinect, using Hidden Markov Model(HMM) and dynamic time warping(DTW) respectively. And the extended and combinational algorithms are realized to improve the accuracy of recognitions. Then systematically analysis of all these algorithms which takes accuracy, complexity and robustness into consideration is done to figure out the best algorithm for the gesture recognition part of the office health system.In addition to drinking and typing, the facial actions and expressions also represent large amount of information during working in office. Therefore, the system of facial movement and expression recognition system is designed and implemented in the thesis. This system can attain diverse areas of facial organs by the data of Face Tracking developer in Kinect for Windows SDK. Then facial movement is achieved by calculating the RGB value in those areas, such as blinking. The system can figure out the working moods of the testers regarding certain facial movements such as laughing, yawning, surprise and sad, and offer suggestions about office heath.