String Actuated Upper Limb Exoskeleton Based on Surface Electromyography Control

This paper discusses a non-pattern recognition method for the exoskeleton implementation. The exoskeleton was controlled using electromyography signal (EMG). The EMG signals were collected at the biceps and processed digitally to control the exoskeleton. The proposed method was a modified low pass filter (LPF) 2nd order using zero crossing (ZC) as a feature extraction. In this study, the proposed method was implemented on a microcontroller using ARM STM32F429 and Discovery board. The output of the model was used to control the exoskeleton using a motor servo. The accuracy was measured using the root mean square error (RMSE) and the Pearson correlation coefficient (CC). In this study, The maximal CC for three varies speed was 0.9856 +/- 0.0012 and the RMSE was 10.30(omicron) +/- 0.484(omicron). This research found that the correlation between the predicted and actual angle was closed. It indicates that the model ZC-LPF could be applied for an upper limb exoskeleton.