機能的電気刺激に基づく筋電位制御インタフェースのトレーニングシステム

抄録

Multifunctional control of EMG-based man-machine interface requires users to generate separated electromyogram (EMG) patterns. This paper proposes a training system to improve the consistency and separability of users’ EMG patterns based on the usage of functional electrical stimulation (FES) and motion selection method. User is capable of contracting target muscles by FES. Motion selection method selects suitable motions (EMG patterns) for each user by eliminating similar ones. Motion classification is performed with a probabilistic neural network using feature patterns extracted from EMG. Experimental results demonstrated that the proposed system is effective for generating separated EMG patterns consistently. It is therefore assumed that proposed system can be used for interface control training.