An estimation model of human dynamic arm strength with joint rotation compensation

Abstract

The measurement of human arm strength is of great significance for human-robot collaboration. As compared with the measurement of human arm strength with force sensors, the estimation of arm strength via electromyogram (EMG) signals is more flexible and convenient, but the corresponding application in industrial field is restricted due to joint rotation and time delay. In order to solve this problem, an estimation model of dynamic arm strength with joint rotation compensation (JRC) is proposed, in which a forgetting parameter is introduced to optimize the feature extraction method, and a deep learning model of joint rotation and EMG signal is constructed to eliminate the impact of joint rotation. Experimental results show that, as compared with the traditional parallel cascade identification (PCI) model, the proposed estimation model is of higher accuracy. Human-robot collaborative sawing experiment indicates that the proposed model can be used as a human-robot collaboration interface.