This paper proposes a novel method for automatically classifying
and retrieving motions based on spatio-temporal features of motion
appearance. Our method first converts a motion data into a form of
clausal language that represents geometrical relations between human
body parts and their temporal relations. A classification rule is then learned from the minimal set of manually classified examples using inductive logic programming. We introduce a two-step search algorithm to retrieve motion segments from the motion database,
which uses two types of classification rules that are discovered by
different learning models. Our method allows robust and efficient
retrieval from complex motion sequences with a small number of
training data.
雑誌名
Proceedings - SCA 2007 : ACM SIGGRAPH / Eurographics Symposium on Computer Animation 2007, Posters and Demos
ページ
22 - 23
発行年
2007-08
出版者
ACM
権利
(C) ACM 2007. This is the author's version of the work. It is posted here for your personal use. Not for redistribution. The definitive Version of Record was published in Proceedings – SCA 2007 : ACM SIGGRAPH / Eurographics Symposium on Computer Animation 2007, Posters and Demos, pp.22-23