Continuous Estimation of a sEMG-Based Upper Limb Joint

The more hidden information from surface electromyography (sEMG) should be extracted for the continuous estimate of the human motion intention based on sEMG. Since feature selection is very important to generalize an estimate model. In this work, the time-domain features (TF), and corresponding time-delayed features (TDF) of sEMG were extracted to estimate human upper limb joint motion. Considering execution time and measure accuracy, Random Forests (RF) algorithm is applied to estimate the joint motion based on the multi-features of sEMG. The difference between the actual angle and the estimated angle were calculated to verify the performance of proposed estimate model. Moreover, the average time of motion estimation is also calculated and the significance of each feature was quantized. Finally, the results showed that the TDF features of sEMG perform well for estimating the joint motion.