A Lightweight Deep Human Activity Recognition Algorithm Using Multiknowledge Distillation

doi:10.1109/JSEN.2024.3443308

	A Lightweight Deep Human Activity Recognition Algorithm Using Multiknowledge Distillation
	Chen, Runze 1; Luo, Haiyong 2; Zhao, Fang 1; Meng, Xuechun 1; Xie, Zhiqing 1; Zhu, Yida 3
	2024-10-01
发表期刊	IEEE SENSORS JOURNAL
ISSN	1530-437X
卷号	24 期号:19 页码:31495-31511
摘要	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.
关键词	Human activity recognition Computational modeling Sensors Feature extraction Training Task analysis Deep learning Artificial neural network human activity recognition (HAR) multiknowledge distillation
DOI	10.1109/JSEN.2024.3443308
收录类别	SCI
语种	英语
资助项目	Strategic Priority Research Program of Chinese Academy of Sciences[XDA28040500] ; National Natural Science Foundation of China[62261042] ; Key Research Projects of the Joint Research Fund for Beijing Natural Science Foundation and Fengtai Rail Transit Frontier Research Joint Fund[L221003] ; Beijing Natural Science Foundation[4232035] ; Beijing Natural Science Foundation[4222034]
WOS研究方向	Engineering ; Instruments & Instrumentation ; Physics
WOS类目	Engineering, Electrical & Electronic ; Instruments & Instrumentation ; Physics, Applied
WOS记录号	WOS:001329294500030
出版者	IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC
引用统计
文献类型	期刊论文
条目标识符	http://119.78.100.204/handle/2XEOYT63/39480
专题	中国科学院计算技术研究所期刊论文_英文
通讯作者	Luo, Haiyong; Zhao, Fang
作者单位	1.Beijing Univ Posts & Telecommun, Sch Comp Sci, Beijing 100876, Peoples R China 2.Chinese Acad Sci, Inst Comp Technol, Res Ctr Ubiquitous Comp Syst, Beijing 100190, Peoples R China 3.Meituan, Chaoyang 100102, Beijing, Peoples R China
推荐引用方式 GB/T 7714	Chen, Runze,Luo, Haiyong,Zhao, Fang,et al. A Lightweight Deep Human Activity Recognition Algorithm Using Multiknowledge Distillation[J]. IEEE SENSORS JOURNAL,2024,24(19):31495-31511.
APA	Chen, Runze,Luo, Haiyong,Zhao, Fang,Meng, Xuechun,Xie, Zhiqing,&Zhu, Yida.(2024).A Lightweight Deep Human Activity Recognition Algorithm Using Multiknowledge Distillation.IEEE SENSORS JOURNAL,24(19),31495-31511.
MLA	Chen, Runze,et al."A Lightweight Deep Human Activity Recognition Algorithm Using Multiknowledge Distillation".IEEE SENSORS JOURNAL 24.19(2024):31495-31511.