TY - GEN
T1 - Enhancing the robustness of EMG-PR based system against the combined influence of force variation and subject mobility
AU - Asogbon Samuel, Mojisola G.
AU - Williams Samuel, Oluwarotimi
AU - Geng, Yanjuan
AU - Idowu, Paul Oluwagbenga
AU - Chen, Shixiong
AU - Ganesh, R. Naik
AU - Feng, Pang
AU - Li, Guanglin
PY - 2018/9/17
Y1 - 2018/9/17
N2 - Inevitable variation in muscle contraction force while performing a target limb movement has been reported to have substantial impact on the performance of electromyography pattern recognition (EMG-PR) based prostheses. The mobility of subject has also been shown to cause changes in the EMG signal patterns when eliciting identical limb movement in mobile scenarios, thus leading to degradation in the overall performance of EMG-PR based prostheses. While the effect of variation in muscle contraction force and subject mobility (VMCF-SM) have only been studied individually, their combined effect on the performance of EMG-PR motion classifier remains unknown. Firstly, we investigated the combined effect of VMCF-SM on the performance of EMG-PR motion classifier by recording EMG signals from five able-bodied subjects in two scenarios (sitting and walking), across low (20% MVC), moderate (50% MVC), and high (80% MVC) muscle contraction force levels. Secondly, we proposed a new time-domain feature set (invTDF) that is robust to VMCF-SM and compared its performance with that of three different widely applied feature extraction methods. The proposed invTDF led to significant reduction in classification error in the range of 6.74% ~ 13.52% with respect to the other feature sets. These preliminary results indicate that using the proposed invTDF may increase the robustness of EMG-based myoelectric control against the combined effect of VMCF-SM.
AB - Inevitable variation in muscle contraction force while performing a target limb movement has been reported to have substantial impact on the performance of electromyography pattern recognition (EMG-PR) based prostheses. The mobility of subject has also been shown to cause changes in the EMG signal patterns when eliciting identical limb movement in mobile scenarios, thus leading to degradation in the overall performance of EMG-PR based prostheses. While the effect of variation in muscle contraction force and subject mobility (VMCF-SM) have only been studied individually, their combined effect on the performance of EMG-PR motion classifier remains unknown. Firstly, we investigated the combined effect of VMCF-SM on the performance of EMG-PR motion classifier by recording EMG signals from five able-bodied subjects in two scenarios (sitting and walking), across low (20% MVC), moderate (50% MVC), and high (80% MVC) muscle contraction force levels. Secondly, we proposed a new time-domain feature set (invTDF) that is robust to VMCF-SM and compared its performance with that of three different widely applied feature extraction methods. The proposed invTDF led to significant reduction in classification error in the range of 6.74% ~ 13.52% with respect to the other feature sets. These preliminary results indicate that using the proposed invTDF may increase the robustness of EMG-based myoelectric control against the combined effect of VMCF-SM.
KW - effect of subjects' mobility
KW - EMG-pattern recognition
KW - muscle contraction force
KW - upper-limb prostheses control
UR - http://www.scopus.com/inward/record.url?scp=85055515873&partnerID=8YFLogxK
U2 - 10.1109/ACIRS.2018.8467236
DO - 10.1109/ACIRS.2018.8467236
M3 - Conference contribution
AN - SCOPUS:85055515873
T3 - Proceedings of 2018 3rd Asia-Pacific Conference on Intelligent Robot Systems, ACIRS 2018
SP - 12
EP - 17
BT - Proceedings of 2018 3rd Asia-Pacific Conference on Intelligent Robot Systems, ACIRS 2018
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 3rd Asia-Pacific Conference on Intelligent Robot Systems, ACIRS 2018
Y2 - 21 July 2018 through 23 July 2018
ER -