No abstract

No abstract

The objective of this paper is to propose a mechanomyogram (MMG)-based motion classification system comprised of a muscle-activity onset detector and a motion classifier. The detector identifies muscle-activity onset time using sampled time-series of MMG signals of biceps and triceps brachii of a human upper arm based on the Mahalanobis-Taguchi method. The classifier is based on the Recognition-Taguchi method and an AdaBoost ensemble learning technique, and distinguishes the flexion and extension of an elbow directly from time-series data of MMG signals of biceps and triceps brachii. We conducted an experimental comparison of the proposed classification system with our previous one based on discriminant analysis techniques to evaluate performance with research participants. Results verified the feasibility of the system and showed that the proposed system achieved higher classification performance than the previous one.