3704 三軸加速度と心電情報を用いた行動判別手法の改善(S55-1 生物医学工学における計測と制御(1),21世紀地球環境革命の機械工学:人・マイクロナノ・エネルギー・環境)

概要

論文の詳細を見る
Recently developed classification schemes of human actions deal with dynamic actions and therefore are not adequate for modeling static actions such as eating and sitting, which are necessary to the monitoring of an entire daily life pattern. In this paper, we propose an enhanced classification method of human actions, that can be applied to both static and dynamic actions. This method combined triaxial accelerometer data and electrocardiogram data. We found that the choice of the part of human body onto which the accelerometer was placed largely affected the classification scores, with wrist being the best candidate for enhanced classification of static actions. The classification scheme used a nearest neighbor algorithm of selected features extracted from triaxial accelerometer data and electrocardiogram data. Fluctuations in the classification results caused by differences in individuals could be canceled using a normalization scheme. Actions such as walking, climbing the stairs, driving, sitting, desk working, eating, standing, going to the bathroom were classified with a score larger than 60%.
社団法人日本機械学会の論文
2008-08-02