<論文>動作プリミティブ抽出と舞踊符割当ての自動化

概要

論文の詳細を見る
We have proposed a human motion description method using Buyo-fu, which aims at establishing a new recording method of human motions and re-using a time series of three-dimensional human motion data(mocap data) obtained by using a motion capturing system. The mocap data can be partitioned into some basic motions. We assign a code to each basic motion, and call it Buyo-fu. We can efficiently describe the human motions by using Buyo-fu. At present, we have a problem that it takes much time and effort to make Buyo-fu, because only a person who has professional knowledge about the concerned motions can extract Buyo-fu, In this paper, we propose a method which can automatically extract Buyo-fu from the mocap data, In our proposed method, 1) a series of the human motions is partitioned into the basic motions based on the speed of the motions, 2) since we consider that similar basic motions appear several times in a series of motions, we do clustering the obtained basic motions, 3) we assign a label to each clustered basic motion, and call it a primitive motion, 4) we can express a series of motions by means of the concatenation of the primitive motions, 5) when we can find out the same pattern of the primitive motions in the human motions, we define this pattern to be a new primitive. Each primitive motion which can be obtained using our proposed method can become Buyo-fu. Performing the experiment with the typical form of "tensho" and "sanchin" of Karate-do, we show the proposed method is a good candidate for automatically extracting Buyo-fu from the mocap data.
秋田大学の論文
2002-10-31

<論文>動作プリミティブ抽出と舞踊符割当ての自動化

スポンサーリンク

概要

著者

関連論文

スポンサーリンク