Skip to main navigation Skip to search Skip to main content

Selection of pose configuration parameters of motion capture data based on dynamic time warping

  • Polish-Japanese Academy of Information Technology

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

1 Citation (Scopus)

Abstract

The paper deals with the problem of motion data classification based on dynamic time warping transform. In the preliminary stage the selection of pose configuration parameters by hill climbing procedure is carried out. To construct dissimilarity matrices Euler angles and unit quaternion distance functions are taken into consideration and compared. To examine proposed approach gait database containing data of 30 different humans is used. The obtained results are satisfactory. The classification has over than 90% accuracy and joint selection allows to reduce dimensionality of pose configuration space noticeably.

Original languageEnglish
Title of host publicationProceedings of the International Conference on Numerical Analysis and Applied Mathematics 2014, ICNAAM 2014
EditorsTheodore E. Simos, Theodore E. Simos, Theodore E. Simos, Charalambos Tsitouras
PublisherAmerican Institute of Physics Inc.
ISBN (Electronic)9780735412873
DOIs
Publication statusPublished - 10 Mar 2015
EventInternational Conference on Numerical Analysis and Applied Mathematics 2014, ICNAAM 2014 - Rhodes, Greece
Duration: 22 Sept 201428 Sept 2014

Publication series

NameAIP Conference Proceedings
Volume1648
ISSN (Print)0094-243X
ISSN (Electronic)1551-7616

Conference

ConferenceInternational Conference on Numerical Analysis and Applied Mathematics 2014, ICNAAM 2014
Country/TerritoryGreece
CityRhodes
Period22/09/1428/09/14

Keywords

  • attribute selection
  • classification of motion data
  • dynamic time warping
  • gait identification
  • motion capture

ASJC Scopus subject areas

  • General Physics and Astronomy

Fingerprint

Dive into the research topics of 'Selection of pose configuration parameters of motion capture data based on dynamic time warping'. Together they form a unique fingerprint.

Cite this