触覚センサを用いた電動義手制御のための動作判別システムに関する研究

抄録

This study proposes a real-time motion classification system for powered prosthetic hand control using forearm surface tactile signals measured by PVDF film sensors. Unlike conventional myoelectric control based on surface electromyography, the proposed approach captures mechanical muscle activity reflected as subtle skin-surface deformations. An offline signal processing and classification framework was first established, including motion onset detection and comparison of BPNN, k-NN, and SVM classifiers. Based on these results, a real-time implementation with a reject option was developed and embedded in a compact hardware system. Experimental results demonstrated high classification accuracy, rapid response, and improved robustness against misclassification, indicating the feasibility of this approach as a novel alternative to conventional myoelectric prosthetic control.