A Lightweight Deep Human Activity Recognition Algorithm Using Multiknowledge Distillation
IEEE Sensors Journal · 01.10.2024 · 24 (19) · 31495-31511
DOI: 10.1109/jsen.2024.3443308 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Sensors Journal, erschienen am 01.10.2024. Verfasst von Runze Chen und 5 weiteren Autor:innen. Die Arbeit wurde bisher 8-mal zitiert (Crossref) und nennt 59 Literaturangaben. Gefördert von: Strategic Priority Research Program of Chinese Academy of Sciences, National Natural Science Foundation of China, Key Research Projects of the Joint Research Fund for Beijing Natural Science Foundation and Fengtai Rail Transit Frontier Research Joint Fund.
Abstract (Originalsprache)
Human activity recognition (HAR) is crucial in fields such as human-computer interaction, motion estimation, and intelligent transportation. Yet, attaining high accuracy in HAR, especially in scenarios limited by computing resources, poses a considerable challenge. This article presents Stage-Memory-Logits Distillation (SMLDist), a framework designed to build highly customizable HAR models that achieve optimal performance under various resource constraints. SMLDist prioritizes frequency-related features in its distillation process to bolster HAR classification robustness. We also introduce an auto-search mechanism within heterogeneous classifiers to boost the performance further. Our evaluation addresses the challenges of generalizing across users, sensor placements, and recognizing a wide array of activity modes. Models crafted with SMLDist, leveraging a teacher-based approach that achieves a 40%–50% reduction in operational expenditure, surpass the performance of existing state-of-the-art architectures. When assessing computational costs and energy consumption on the Jetson Xavier AGX platform, SMLDist-based models show strong economic and environmental sustainability advantages. Our results indicate that SMLDist effectively alleviates the performance degradation typically associated with limited computational resources, underscoring its significant theoretical and practical contributions to the field of HAR.
Förderung
- Strategic Priority Research Program of Chinese Academy of Sciences
- National Natural Science Foundation of China
- Key Research Projects of the Joint Research Fund for Beijing Natural Science Foundation and Fengtai Rail Transit Frontier Research Joint Fund
- Beijing Natural Science Foundation
Zitieren
APA
Chen, R., Luo, H., Zhao, F., Meng, X., Xie, Z., & Zhu, Y. (2024). A Lightweight Deep Human Activity Recognition Algorithm Using Multiknowledge Distillation. IEEE Sensors Journal, 24(19), 31495-31511. https://doi.org/10.1109/jsen.2024.3443308
Vancouver
Chen R, Luo H, Zhao F, Meng X, Xie Z, Zhu Y. A Lightweight Deep Human Activity Recognition Algorithm Using Multiknowledge Distillation. IEEE Sensors Journal. 2024;24(19):31495-31511. doi:10.1109/jsen.2024.3443308
BibTeX
@article{Chen2024,
author = {Chen, Runze and Luo, Haiyong and Zhao, Fang and Meng, Xuechun and Xie, Zhiqing and Zhu, Yida},
title = {A Lightweight Deep Human Activity Recognition Algorithm Using Multiknowledge Distillation},
journal = {IEEE Sensors Journal},
year = {2024},
volume = {24},
number = {19},
pages = {31495--31511},
doi = {10.1109/jsen.2024.3443308}
}
EUVIMED