2-209 知的運動補助装置の構築における個人特性に影響を及ぼす要因の分析 : 筋力変化が EMG 動作識別器に与える影響

概要

論文の詳細を見る
Learning of bio-signal-to-motion classifiers would be influenced by the factors like muscular forces and postures of the motion-related body part. In this paper, we analyzed the influence of muscular force on individual characteristics (that is means map relation between a signal and discriminating motion) using a bio-signal information criterion, which is based on the concept of conditional information entropy. This criterion was possible to reflect more correctly the degree of complexity between clusters formed by bio-signal data compared with other criterion like DB measure and KL information criterion. An experiment on forearm motion discrimination under five muscular force levels (based on 10,20,30,50,70%MVC) was conducted to verify the influence. Results indicated that change in muscular force influences individual characteristics. In the case of muscular force with 10,50%MVC, it was found through evaluation by entropy that muscular force is not enough, since information for discriminating motion was less. Consequently, we concluded that the muscular force with 20 or 30%MVC was suitable, and it was important to feed back their muscular force level to user in order to abolish individual variation.
一般社団法人日本機械学会の論文
2002-11-14