Motion recognition based on temporal mode acquisition

Abstract

In this paper, a method for constructing ideal motion and analyzing difference between current and past motion is proposed. The ideal motion means a motion which is considered from past to current motion information. The proposed method is supposed to apply motion assist by robots. For the motion assist by robots, individual database adjusted to day-by-day variation is necessary. This database should be able to extract ideal motion in daily motion and acquire the difference between current and past motion. By analyzing difference between current and past motion, robots can recognize day-by-day variation of the examinee. For deriving ideal motion and analyses of the difference at the same time, quarry matrix is used. Finally, this paper generates the ideal motion considering the correlations between past ideal motion and the elements in the database. To recognize both ideal motion and differences, this paper extracts the force information using temporal mode. The proposed method enables to analyze the difference between elements in the database without impairing the correlations. The viability of the proposed method is confirmed by experiments.