| In the field of modern rehabilitation,arm motion recognition as an important work of intelligent rehabilitation,it has attracted extensive attention of researchers.If the rehabilitation training patterns of patients with limb dysfunction can be identified,and rehabilitation training strategies can be customized for them,and the recognition algorithm is combined with virtual reality technology to build a rehabilitation training system,modern rehabilitation training will be more efficient and intelligent.In this paper,theory and experiment on the action recognition of arm is carried out.The research contents are as follows:Firstly,the concept of equality constraint is introduced into kalman filter with random error of Kinect,and the human body skeleton features with constant length is applied to the human body skeleton topology structure,then a restrictive Kalman filtering algorithm with equality constraints is proposed,and the mathematical model of improved Kalman filter is established.the concept of convolution moving average is introduced with large local error of joint range,and an adaptive convolution moving average filter is proposed,which is based on the reliability of the filtering results.Secondly,for obtaining the continuity equation of human joint motion,a motion calculation method of human joint activity based on joint space sequence is proposed based on obtain the topological structure of human skeleton with Kinect camera,and the factors affecting the accuracy of joint activity are analyzed,and a rehabilitation information system using Kinect is designed.With the problem of doctor-patient occlusion,a human joint point repair model,which is suitable for human skeleton calibration,is established based on the characteristics of human kinematics constraints.Thirdly,for the personalisation and diversification of users,the design flow and processing flow of program are designed in detail,the software and hardware of arm motion recognition is developed,and the compatibility and rationality of the system are optimized;Then,the human-interaction interface is built,and the real-time test experiment is carried out.Finally,for verifying the stability and accuracy of the measurement system,the static and dynamic experiments of arm motion recognition system are carried out.The error of the measurement system is obtained based on the experiment data.The consistency of the experiment results is verified by hypothesis test,and the source of the error is analyzed;At the same time,a human-machine synchronous motion control method is designed for biosyncretic elbow and wrist rehabilitation robot,and a new remote rehabilitation training mode is proposed. |
