Reconstruction of Human Motions Using Few Sensors
Abstract
In this paper we show that it is possible to reconstruct whole body motions from the data taken by as few as 1, 2, 4 inertial sensors, if a semantic pre-classification of the motion is available, and motion data bases can be used to synthesize missing degrees of freedom. We demonstrate some examples of reconstructed motions and compare them with a video of the take, and we use data of optical motion captures for numerical comparisons. Although our work is mainly inspired by the wish to create easy-to-use and cheap motion capture systems for everyday settings, such a motion capture system is also an easy-to-use tracking system of whole body motions. Moreover, the underlying synthesis of motion out of low-dimensional control signals opens new ways for the control of avatars.
Best Paper Award
Images
![]() |
Download Paper
Additional Material
- Video (AVI video, 9.2 MB)
Bibtex
@INPROCEEDINGS{tautges-2008-reconstruction, author = {Tautges, Jochen and Kr{\"u}ger, Bj{\"o}rn and Zinke, Arno and Weber, Andreas}, editor = {Schumann, M. and Kuhlen, T.}, pages = {1--12}, title = {Reconstruction of Human Motions Using Few Sensors}, booktitle = {Virtuelle und Erweiterte Realit{\"a}t: 5. Workshop der GI-Fachgruppe VR/AR}, year = {2008}, month = sep, publisher = {Shaker Verlag}, location = {Magdeburg, Germany}, abstract = {In this paper we show that it is possible to reconstruct whole body motions from the data taken by as few as 1, 2, 4 inertial sensors, if a semantic pre-classification of the motion is available, and motion data bases can be used to synthesize missing degrees of freedom. We demonstrate some examples of reconstructed motions and compare them with a video of the take, and we use data of optical motion captures for numerical comparisons. Although our work is mainly inspired by the wish to create easy-to-use and cheap motion capture systems for everyday settings, such a motion capture system is also an easy-to-use tracking system of whole body motions. Moreover, the underlying synthesis of motion out of low-dimensional control signals opens new ways for the control of avatars.}, isbn = {978-3-8322-7572-3}, conference = {5. Workshop "Virtuelle und Erweiterte Realit\"{a}t" der GI-Fachgruppe VR/AR} }