Research And Implementation Of Knee Joint Exercise Rehabilitation System Based On Double Kinect Data Fusion

The wrong way of movement and unhealthy living posture lead to the prevalence of knee joint problems among the general public.In order to deal with the problems of traditional rehabilitation medical methods,such as high cost,lack of medical resources and medical difficulties caused by unexpected situations,in this thesis,we design and implement a system based on dual Kinect somatosensory equipment,which combines data fusion technology and human action posture recognition technology,aiming at providing professional knee rehabilitation training guidance for patients.This thesis expands the vision of single Kinect equipment by dual Kinect equipment,solves the limitation of sensing range existing in single device,and analyzes the relationship between the actual physical placement position and effective sensing range of dual devices through experiments.Finally,the most appropriate physical placement position of dual Kinect devices was analyzed and determined.Experiments show that the placement method based on the distance between two devices of 2M and the angle of 90 ° is not only suitable for human rehabilitation training It provides enough activity space and is suitable for most users' home environment,and ensures the integrity of data acquisition.In view of the existing problem of single Kinect equipment,such as bone jitter caused by random noise error of sensors,data loss caused by strong light pollution or object occlusion environment and data deviation caused by self-occlusion caused by standing position of human body relative to camera,this thesis optimizes the data fusion processing of human bone data from different perspectives obtained by dual Kinect equipment,and optimizes the data fusion processing through the constraints of human physiological and motion characteristics.The reliability and stability of human bone data can be improved by filtering trusted data and optimizing the weight distribution of fusion formula according to the contribution degree of human bones.The improved Kalman filter is used to make up for the missing bone points of human nonlinear motion state in real time.In human action recognition,the effect of traditional template matching recognition is affected by the difference of individual action position and template action position.In order to solve this problem and improve the recognition error caused by the difference of height and weight of different individual users,in this thesis,based on the credible human skeleton point coordinates obtained by dual Kinect data fusion,extract the angle between bones and the spatial position relationship between bone points corresponding to the main active joint points of action posture,and normalize them.Combined with multi feature constraints,the traditional template matching algorithm is improved,which improves the recognition effect of human knee exercise rehabilitation posture,and provides strong support for the design and implementation of knee exercise rehabilitation system based on dual Kinect data fusion.