Research And Design Of Human Rehabilitation System Based On Kinect

Nowadays the number of patients with disability or paralysis increases rapidly, so the demand of rehabilitation training becomes more urgent. At present the medical rehabilitation mainly focuses on artificial assisted rehabilitation and robot assisted rehabilitation, but due to the shortage of health care providers, the high cost of intelligent robots and its small-scale applicability etc, artificial and robot assisted rehabilitation requires great manpower and material resources. With the rapid development of the computer vision technology, technology based on image analysis is widely used in various fields, such as industrial product detection, agricultural pest control, traffic monitoring, visual management and visualization of medical care etc. Computer vision technology costs less and becomes more mature, so it is of great significance to design a medical rehabilitation training system based on computer vision technology, so as to meet the medical needs of a large number of patients with disease.In this thesis, we put computer vision technology as the basis to solve the problem, and use the Microsoft Kinect to obtain skeletal information about training of the patients to research and design a human rehabilitation system based on Kinect. The system designs a variety of training scenes and training tasks. Under the guidance of the rehabilitation coaches and doctors, patients choose training scenes and complete the training task, then the system evaluates it and returns the results to the doctor to monitor the progress of rehabilitation training. The main research contents are as follows:(1) Study on the method of 3D scene rendering. Based on Levels of Detail (LOD) terrain rendering technology and SilverLining sky rendering technology, a combination of 3D scene rendering technology is proposed, which is used to design colorful rehabilitation training scenes.(2) Study on the method of evaluating the effect of rehabilitation training. Based on the combination of Longest Common Subsequence (LCS) algorithm and RANSAC algorithm, a dual algorithm, LCS-RANSAC. is proposed for similarity assessment of action stream. It effectively solves the evaluation problem of rehabilitation training for patients, and provides reliable medical basis to rehabilitation coaches and doctors.(3) Study on the method of rehabilitation training. A rehabilitation task training method based on the acquisition technology of Kinect skeleton information is proposed. The method controls the character model in real time by skeletal information to complete training tasks, and acquires standard action stream of rehabilitation coaches and training action stream of patients to achieve rehabilitation training by contrasting and evaluating those two streams.