20-7 情報セラピーインタフェースのための画像による振る舞い検出手法の検討(第20部門 ヒューマンインフォメーション)

抄録

This paper describes a method for assisting human activities based on human posture estimation using sparse 3-D depth observations. In the proposed method, human behavior is modeled as transitions of 3-D appearances, and the results of this behavior detection are used to determine instructions to be given to the user. We show an sample implementation of the system for a toilet task.