A Stretch Rehabilitation Training System Based On Kinect

In the field of medical rehabilitation,the two most important methods are artificial rehabilitation and machine assisted rehabilitation.Because of the simplex function,high equipment cost and maintenance cost,the market of machine assisted rehabilitation is diamal.Artificial rehabilitation is effective and has good interactivity.But the number of patients is far greater than the number of rehabilitation coaches.This unequal relationship between supply and demand led to the extreme shortage of artificial rehabilitation resources.In view of the high cost and difficult maintenance of machine rehabilitation and lack of resources of artificial rehabilitation,to provide a rehabilitation system which can not only adapt to the ability to buy but also can ease lack of resources of artificial rehabilitation is an important issue in the field of medical rehabilitation.In order to solve the above problems,we aim to develop a convenient and scientific rehabilitation training system that can help patients train at home or in the company.The system is based on the concept of human-computer interaction.Microsoft Kinect is the hardware foundation.Detecting the similarity between patient actions and standard actions is the main content of the article.The main work of this thesis is as follows:(1)Introducing the basic hardware architecture and software development environment of Kinect and the overall system architecture of rehabilitation training system.This thesis introduces the logical structure and system data flow of the whole system.The whole system consists of patients,Kinect equipment and professional medical staff.First of all,the medical staff establish a standard rehabilitation training action bank which exists in the system according to the professional medical knowledge.Patients train according to the standard action and Kinect capture their actions.Then the system does data processing and analysises the similarities of the two actions and gives feedback to the user.(2)Research on the method of action similarity matching.In this thesis,the static gesture recognition algorithm is based on angle measurement.Dynamic time warping(Dynamic Time Warping DTW)algorithm is used for action similarity matching.Random sample consensus(Random Sample Consequence RANSAC)algorithm helps verify the accuracy of the result of DTW algorithm.Standard action library was established first.Then Kinect captures the depth data and constructs human skeleton structure.The system does action similarity matching based on the sum of the Euclidean distances between the frames of the two action streams and gives feedback to the users.(3)Proof and accuracy analysis of algorithm.For the static gesture recognition algorithm,three static postures are defined.For each pose,several different characters are made to verify the recognition results several times.For the action flow similarity matching algorithm,three standard action streams are set so that different people can verify the accuracy of the algorithm.The results and the accuracy of the experiments are analyzed.(4)Detailed design and implementation of the whole stretch rehabilitation training system,including the human skeleton data stream storage method,data acquisition,data processing,design and contact of each module of the system,character Avator of XNA Game Studio platform and the usage of the whole system.