時系列性を生かした表面筋電位に基づく手指の運動推定

抄録

Methods to estimate continuous finger motion from EMG considering its time series features are examined. Although EMG in continuous motion seems to include dynamic properties, conventional methods utilizes EMG feartures only statically. In this presentation, to utilize the dynamic properties, an artificial neural network with time delays and one with a recursive structure are examined. The validity of the both metods are confirmed by experiment. As a result, the higher accuracy is obtained for unknown data sets comparing to a conventional method.