Decoding sEMG Under Non-Ideal Conditions Toward Robust Muscle-Machine Interface Control

• Published in: 2023 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) pp. 4115 - 4120
• Main Authors: Guerrero-Mendez, C.D., Blanco-Diaz, C.F., Lopez-Delis, A., Bastos-Filho, T., Andrade, R.M.
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.10.2023
• Subjects: Extreme learning machines; Fatigue; Finance; Machine learning; Machine learning algorithms; Muscles; Scholarships
• ISSN: 2153-0866
• DOI: 10.1109/IROS55552.2023.10341503

The evaluation of systems under non-ideal conditions is a research problem, particularly in robotic applications for the rehabilitation of people with disabilities. Accordingly, the evaluation of algorithmic strategies for robustness validation under different non-ideal conditions is a current challenge for the scientific community. Therefore, in this study, a computational methodology based on Extreme Learning Machine (ELM) was evaluated for the recognition of seven hand gestures using sEMG under five non-ideal conditions. The shift of eight sEMG electrodes, three upper-limb postures, increased muscle fatigue, and inter-subject and inter-day variabilities were evaluated. The results indicate that the proposed methodology performs well under specific conditions in comparison with previous strategies reported in the literature using Machine Learning classifiers. Therefore, the findings of this study are potentially important for the field of robotics; however, more efforts are still needed to develop more robust computational methods to obtain higher accuracy under non-ideal conditions, with the aim of implementing more controllable, usable, and reliable systems.