Comparing of Electromyography and Ultrasound for Estimation of Joint Angle and Torque.

Electromyography (EMG) signals and ultrasound sensing have been widely studied to estimate movement intentions in the human-machine interface (HMI) field. In this study, we compared the EMG and ultrasound for estimation of the positions and torque of the wrist joint. The estimation performance of EMG and ultrasound were compared and analyzed. EMG, ultrasound, and torque signals were collected from 8 able-bodied subjects. The subjects were instructed to conduct isometric contraction of the wrist flexion/extension at 13 different angular positions. Linear regression (LR) was adopted to estimate isometric contraction torque while support vector machine (SVM) was employed to identify different angular positions. Regarding isometric contraction torque estimation, the average Pearson’s correlation coefficient (r) values across all subjects was 0.95 and 0.84 for EMG and ultrasound, respectively. Regarding angular position classification, the average classification accuracy (CA) across all subjects of EMG and ultrasound was 17.78% and 91.05%, respectively. The results demonstrated that the EMG outperformed the ultrasound in estimating torque while the ultrasound outperformed the EMG for angular position classification. Our study demonstrated the advantages of the different types of signals to achieve the estimation of the joint angle and torque.