Surface EMG based upper limb motion recognition in real-time

Through the sEMG signal from patients' healthy limb to control the robot and assist their hemiplegic limb to train is a promising rehabilitation method. For this purpose, sEMG based upper limb motion recognition method in real-time was proposed using four time domain features and Support Vector Machine method. The continuity of arm movement requires the sEMG based robot control to be continuous. Therefore, a motion data partition strategy was proposed: Firstly determine the appropriate partition interval by experiment according to the classification accuracy and then take the accuracy and continuity into consider, using the sliding window to segment the data. Set the partition interval found previously as window width, and set the sliding distance with a smaller value. The experiment results show that the method proposed is suitable for real-time application in accuracy and continuity. Copyright © 2013 Binary Information Press.