Research On The Active Intention Perception Technology Of Human For The Upper Limb Rehabilitation Robot

With the aggravation of population aging and frequent occurrence of traffic accidents,the number of patients with motor dysfunction has been increasing day by day.Scientific rehabilitation training is of great value for the recovery of limb movement dysfunction.However,due to the lack of rehabilitation medical institutions and rehabilitation physicians,a large number of patients can't get the necessary rehabilitation training,which leads to the complete loss of limb movement function.The rapid development of robot technology provides a new way to solve rehabilitation problems.Robots can carry out rehabilitation training for a long time.The rehabilitation robot was also born to reduce the workload of rehabilitation physicians,thus increasing the number of rehabilitation training receivers.Rehabilitation robot rehabilitation training is divided into two types:active training and passive training.Passive training robot guide the movement of patients according to the prescribed actions.The implementation of this kind of robot is relatively simple.The active training process needs to perceive the movement intention of the human body,which causes the rehabilitation robot to follow the human body's intention to control the movement.The accurate perception technology of the movement intention of the body has become the difficult point of the active rehabilitation technology.This paper focuses on the active intention perception technology of upper limb rehabilitation robot.First,a kind of exoskeleton upper limb rehabilitation robot was designed and processed.Then,in order to ensure the accurate understanding of the intention and prevent the misunderstanding of the human movement information,this paper studied the selection of the motion features of the EMG signal and improved the recognition accuracy of the muscle electrical signal itself.Finally,a security perception framework based on visual feedback is proposed to avoid the catastrophic consequences of misjudgement.The main contents of this paper are as follows:(1)Construction of the platform for the rehabilitation of the upper limbAccording to the requirements of upper limb rehabilitation training and the size of human upper limbs,the exoskeleton structure of six degree of freedom upper limb rehabilitation robot is designed.Among them,shoulder flexion extension,shoulder joint adduction and external swing,elbow flexion extension,wrist flexion extension were driven directly by driving mechanism.Because the driving device of the exoskeleton robot revolving part needs to be placed outside the robot,the internal rotation/external rotation of the upper arm and the internal rotation/external rotation of the forearm are fixed by three points coplanar with a gear guide wheel and two guide wheels to ensure the movement coaxial of the robot and the human body.Subsequently,ANSYS and Adams were used to complete the static simulation and kinematics simulation respectively,which ensured the safety and reasonable movement of the robot structure.After that,the servo motor,reducer and electrical control system of the upper limb rehabilitation robot are designed and selected.On this basis,the robot is processed and assembled,which validated the design of rehabilitation robot.Finally,the PC's motion control software is written by MFC,and the robot is controlled according to the rehabilitation requirements to achieve passive rehabilitation training.(2)Research on perception of upper limb movement intention based on muscle electrical signalsThe EMG signal can directly reflect the movement intention of the human body.This article collects the EMG signals of the muscles of the biceps brachii,the triceps brachii,the brachial radial muscle and the flexor of the ulnar wrist.The features of the time domain log statistics,the log statistic characteristics after wavelet packet transformation(WPT)and the log statistic characteristics after the empirical mode decomposition(EMD)are used as eigenvectors to characterize the movement pattern of the upper limb.The features input into the neural network classifier of the BPNN to classify the pattern.By studying and classifying the 8 daily movements of the upper limbs,96.7%accuracy is achieved.In order to improve the classification accuracy and speed up the classification speed,the binary particle swarm algorithm(BPSO)and the improved discrete firefly algorithm(IDFA)are used to select the classification features,and the original 192 characteristics are reduced to 89.By analyzing the characteristics of EMG signals in upper limb movements,this article believes that higher frequency of EMG signals can better represent the pattern of limb movement.(3)Research on secure active sensing technology based on visual feedbackBecause of the complexity of the EMG signal itself and the influence of the external interference,the accuracy of intention based on the EMG signal is limited.In the rehabilitation training,the wrong action is completely unacceptable,which leads to the difficult of the active rehabilitation training.In this case,the identification result of the motion of the EMG signal is displayed by the virtual animation to the patient.The patient uses the head action to make confirmation of the action,thus ensuring the safety of the rehabilitation training.The specific processing methods are as follows:the patient nods his/her head to express the agreement of the identification result or shakes his/her head to express the rejection of the identification results.The robot uses Kinect v2.0 to collect the video action of the patient's head.First,the location of the patient's head is determined by AdaBoost's face location method to lock the region of interest(ROI).Then the optical flow in the ROI was extracted from the motion of video.The identification results of the head movement are given by voting,Sigmoid decision or tanh decision,which completes the feedback recognition process.In order to ensure the effectiveness of decision-making,the parameters of decision algorithm are optimized in the end.Through the research of the above part,this paper has completed the perception of the active intention of the upper limb rehabilitation robot.Through the active perception rehabilitation training experiment,it has proved that the method in this paper has a good effect and provides a new way for the active rehabilitation training.