Robust finger motion classification using frequency characteristics of surface electromyogram signals

Abstract

Finger motion classification using surface electromyogram (EMG) signals is currently being applied to myoelectric prosthetic hands with methods of pattern classification. It can be used to classify motion with great accuracy under ideal circumstances. However, the precision of classification falling to change the quantity of EMG feature with muscle fatigue has been a problem. We addressed this problem in this study, which was aimed at robustly classifying finger motion against changes in EMG features with muscle fatigue. We tested the changes in EMG features before and after muscle fatigue and propose a robust feature that uses a methods of estimating tension in finger motion by taking muscle fatigue into consideration.