Automatic Viewpoint Selection for Interactive Motor Feedback Using Principal Component Analysis

Florian Diller, Alexander Wiebel, Gerik Scheuermann

2024

Abstract

We present a novel method to automatically select a viewpoint optimized for the interactive display of a physical exercise which is shown using a human skeleton-like avatar with additional visual motor feedback. Expressive viewpoints are crucial for the users to be able to understand and interactively adapt to the feedback in all its spatial aspects. Selecting camera perspectives for these viewpoints can be challenging when the presentation includes specific visual feedback cues in addition to the instantaneous pose, as many different requirements have to be taken into consideration in this case. The users continuously correcting their movements according to the visual real-time feedback represents a special case of human-computer interaction. Our algorithm employs principal component analysis (PCA) to determine informative viewing directions for the overall pose and specific feedback cues shown at different joints. The final viewpoints are synthesized from the obtained directions in a per-frame manner. To evaluate our method we conducted a user study with 39 participants. They were asked to choose from four exercise videos with motor feedback generated by the presented method and three competing existing approaches. Additionally, to validate our approach’s assumptions, we asked the participants to freely choose a viewpoint, which they considered optimal for the provided motor feedback. The results of the study show that our algorithm was most frequently chosen as being the most informative. Furthermore, our method proved much faster than previous viewpoint selection methods, as it does not require information about upcoming frames. This makes our approach most suitable for real-time and interactive applications.

Paper Citation

in Harvard Style

Diller F., Wiebel A. and Scheuermann G. (2024). Automatic Viewpoint Selection for Interactive Motor Feedback Using Principal Component Analysis. In Proceedings of the 19th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications - Volume 1: HUCAPP; ISBN 978-989-758-679-8, SciTePress, pages 350-361. DOI: 10.5220/0012308700003660

in Bibtex Style

@conference{hucapp24,
author={Florian Diller and Alexander Wiebel and Gerik Scheuermann},
title={Automatic Viewpoint Selection for Interactive Motor Feedback Using Principal Component Analysis},
booktitle={Proceedings of the 19th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications - Volume 1: HUCAPP},
year={2024},
pages={350-361},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0012308700003660},
isbn={978-989-758-679-8},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 19th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications - Volume 1: HUCAPP
TI - Automatic Viewpoint Selection for Interactive Motor Feedback Using Principal Component Analysis
SN - 978-989-758-679-8
AU - Diller F.
AU - Wiebel A.
AU - Scheuermann G.
PY - 2024
SP - 350
EP - 361
DO - 10.5220/0012308700003660
PB - SciTePress