Ryu Katoh, Daisuke Nishikawa, Wenwei Yu, Hiroshi Yokoi and Yukinori Kakazu

Abstract:
We propose evaluation criteria for processing information extraction in discriminating motion from EMG signals. The concept involves conditional information entropy with attention to the ambiguity of information on discriminating motion. Experiments indicated a strong correlation between entropy and the discrimination rate. We evaluated the setting of signal measurement points from proposed evaluation criteria and found that measurement points could be selected.

Keywords: EMG, conditional information entropy, SOM, pattern recognition, and multichannel signal processing