Study On A Surface EMG Based Upper-limb Rehabilitation Robotic System

The development of robot technology provides a good opportunity and combined withclinical rehabilitation medicine rehabilitation robot research. Technology for upper limbrehabilitation robot system is the more recent technology, the use of robots and relatedtechnologies for monitoring and evaluation of the objective of the training process, improvehemiplegia rehabilitation training pertinence and science, at the same time the treatingphysicians out from heavy physical labor in solution, rehabilitation plan for patients todevelop better, further to improve the efficiency of rehabilitation.This paper aims to build one kind of exoskeleton intelligent rehabilitation system,which can implement5degree-of-freedom (DOF) movements including shoulders, elbowsand wrist for bilateral rehabilitation which has been proven to be an effective way fornervous system rehabilitation. The affected limb is driven by the machine to copy thehealthy limb’s movements. The joint angle collected by inertial sensor called MTx from thehealthy limb was viewed as the motion control command. Due to the advantages inunderstanding and revealing the potential movements in muscle, surface EMG sensor wasintroduced to replace the MTx sensor. This paper proposed the Ensemble Empirical ModeDecomposition(EEMD) method and general entropy with parameters to reveal the motioninformation embedded in EMG signal. In contrast with MTx data, the proposed algorithmcould achieve an average82%recognition accuracy rate in continuous elbow movement,confirming the EMG signal could be able to use in the machine control and also theeffective muscle strength evaluation. In order to visualization the rehabilitation process,Kinect was used as visual sensing to build a three-dimensional model of the human bodythrough the skeleton tracking methods. Our visual system could achieve effectiverecognition of human limb movements and also the automatic scoring of upper-limbrehabilitation conditions with Fugl-Meyer assessment criteria. RGB-D features was used tocalculate the upper limb joint position in three-dimensional space as the same recognitionaccuracy rate as EMG did and4%automatic scoring errors for single task and6%for thewhole task with1%variance. It proved effectiveness to apply visual sensing technology inthe rehabilitation system.