PaperarXiv:2311.15831
Temporal Action Localization for Inertial-based Human Activity Recognition
Proc. ACM IMWUT 8(4), Article 174, November 2024
TL;DR
First systematic transfer of single-stage video TAL models — ActionFormer, TriDet (trident boundary head) and TemporalMaxer — onto raw inertial windows (and, optionally, features from a prepended inertial net). Models emit a label plus start/end offsets instead of classifying fixed windows. Benchmark: LOSO, 1 s windows with 50% overlap, six HAR sets including WEAR and Hang-Time (both 50 Hz).
- Form factor: multi
Abstract
As of today, state-of-the-art activity recognition from wearable sensors relies on algorithms being trained to classify fixed windows of data. In contrast, video-based Human Activity Recognition, known as Temporal Action Localization (TAL), has followed a segment-based prediction approach, localizing activity segments in a timeline of arbitrary length. This paper is the first to systematically demonstrate the applicability of state-of-the-art TAL models for both offline and near-online Human Activity Recognition (HAR) using raw inertial data as well as pre-extracted latent features as input. Offline prediction results show that TAL models are able to outperform popular inertial models on a multitude of HAR benchmark datasets, with improvements reaching as much as 26% in F1-score. We show that by analyzing timelines as a whole, TAL models can produce more coherent segments and achieve higher NULL-class accuracy across all datasets. We demonstrate that TAL is less suited for the immediate classification of small-sized windows of data, yet offers an interesting perspective on inertial-based HAR – alleviating the need for fixed-size windows and enabling algorithms to recognize activities of arbitrary length. With design choices and training concepts yet to be explored, we argue that TAL architectures could be of significant value to the inertial-based HAR community. The code and data download to reproduce experiments is publicly available via github.com/mariusbock/tal_for_har.
Verbatim from arXiv (opens in a new tab) · arXiv metadata (abstracts are released under CC0)
Results on ChatPPG leaderboards
All leaderboardsNo results from this paper on ChatPPG leaderboards yet. Browse the leaderboards.
Tasks & signals
Tasks
Signals
Keywords
Datasets used
How it works
Cite this paper
Bock, M., Moeller, M., & Van Laerhoven, K. (2024). Temporal Action Localization for Inertial-based Human Activity Recognition. Proc. ACM IMWUT 8, , November 2024, 174. https://doi.org/10.1145/3699770
Bock, Marius, Michael Moeller, and Kristof Van Laerhoven. 2024. "Temporal Action Localization for Inertial-based Human Activity Recognition." Proc. ACM IMWUT 8, , November 2024, 174. https://doi.org/10.1145/3699770.
Bock, M., Moeller, M. and Van Laerhoven, K. (2024) 'Temporal Action Localization for Inertial-based Human Activity Recognition', Proc. ACM IMWUT 8, , November 2024, p. 174. doi: 10.1145/3699770.
Bock, Marius, et al. "Temporal Action Localization for Inertial-based Human Activity Recognition." Proc. ACM IMWUT 8, , November 2024, 2024, p. 174. https://doi.org/10.1145/3699770.
M. Bock, M. Moeller, and K. Van Laerhoven, "Temporal Action Localization for Inertial-based Human Activity Recognition," Proc. ACM IMWUT 8, , November 2024, p. 174, 2024, doi: 10.1145/3699770.
@inproceedings{bock2024temporal,
title = {{Temporal Action Localization for Inertial-based Human Activity Recognition}},
author = {Bock, Marius and Moeller, Michael and Van Laerhoven, Kristof},
year = {2024},
booktitle = {Proc. ACM IMWUT 8, , November 2024},
pages = {174},
doi = {10.1145/3699770},
eprint = {2311.15831},
archivePrefix = {arXiv}
}Added to the ChatPPG research hub