EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Deployment-Oriented Evaluation of Lightweight IMU-Based Human Activity Recognition: On-Device Efficiency and Streaming Feasibility

Inho Gil iD, Seongmin Ha iD, Jihwan Oh iD, Donggyu Lee, Soonwoong Hwang, Joonkyu No iD, Wansoo Kim iD

Sensors · 2026

DOI: 10.3390/s26175678 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Sensors, 2026. Verfasst von Inho Gil und 6 weiteren Autor:innen. Beteiligte Einrichtung: Department of Robotics, Hanyang University, 222 Wangsimni-ro, Seongdong-gu, Seoul 04763, Republic of Korea. Hauptthemen laut MEDLINE: Human Activities, Wearable Electronic Devices. Die Arbeit nennt 31 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Ministry of Health and Welfare, Assistive Technology Commercialize R&D Project for Independent Living for People with Disability and Older People, National Research Foundation of Korea.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

On-device Human Activity Recognition (HAR) requires balancing accuracy and deployment efficiency on constrained hardware. We present Lightweight Human Activity Recognition (L-HAR), a controlled comparison of Baseline and Lightweight Temporal Convolutional Network (TCN), Transformer, and Long Short-Term Memory (LSTM) models using data sampled at 100Hz from three inertial measurement units (IMUs) worn by one participant. Evaluations covered classification, model footprint, multiply–accumulate operations (MACs), offline CPU latency, Desktop/Raspberry Pi streaming, 8-bit integer (INT8) quantization, and software-estimated Raspberry Pi energy efficiency. Lightweighting reduced parameters by up to 95.5%, model size by 94.0%, and MACs by 87.6–95.5%, with accuracy reductions of 0.8–2.2 percentage points (pp) and sub-millisecond offline latency for all Lightweight models. Baseline and Lightweight models sustained approximately 100Hz message/inference rates on Desktop, whereas no Raspberry Pi configuration reached 100Hz inference. Among Lightweight models, Transformer achieved 9.381ms End-to-End (E2E) latency and 66.925Hz inference. INT8 quantization changed accuracy and F1-score by less than 0.13 pp; Quantized TCN achieved the best Raspberry Pi streaming result (9.066ms E2E; 69.928Hz) with an estimated 49.933mJ per inference. These results demonstrate architecture- and backend-dependent deployment behavior and the need for direct target-platform evaluation.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

Human ActivitiesWearable Electronic DevicesHumansLong Short Term MemoryAlgorithmsConvolutional Neural Networks

Stichwörter der Autor:innen

LSTMTCNdeep learning modelshuman activity recognitioninertial measurement unitslightweighton-device efficiencytransformerwearable sensors

Autor:innen und Institutionen

  • Department of Robotics, Hanyang University, 222 Wangsimni-ro, Seongdong-gu, Seoul 04763, Republic of Korea.
  • Department of Mechatronics Engineering, Hanyang University, 222 Wangsimni-ro, Seongdong-gu, Seoul 04763, Republic of Korea.
  • Department of Robotics, Hanyang University ERICA, 55 Hanyangdaehak-ro, Sangnok-gu, Ansan 15588, Republic of Korea.

ORCID-Profile: Inho Gil, Seongmin Ha, Jihwan Oh, Joonkyu No, Wansoo Kim

Förderung

  • Ministry of Health and Welfare
  • Assistive Technology Commercialize R&D Project for Independent Living for People with Disability and Older People
  • National Research Foundation of Korea
  • Institute of Information & Communications Technology Planning & Evaluation (IITP) under the Artificial Intelligence Innovation Human Resources Development

Zitieren

APA

Gil, I., Ha, S., Oh, J., Lee, D., Hwang, S., No, J., & Kim, W. (2026). Deployment-Oriented Evaluation of Lightweight IMU-Based Human Activity Recognition: On-Device Efficiency and Streaming Feasibility. Sensors. https://doi.org/10.3390/s26175678

Vancouver

Gil I, Ha S, Oh J, Lee D, Hwang S, No J, et al. Deployment-Oriented Evaluation of Lightweight IMU-Based Human Activity Recognition: On-Device Efficiency and Streaming Feasibility. Sensors. 2026. doi:10.3390/s26175678

BibTeX
@article{Gil2026,
  author = {Gil, Inho and Ha, Seongmin and Oh, Jihwan and Lee, Donggyu and Hwang, Soonwoong and No, Joonkyu and Kim, Wansoo},
  title = {Deployment-Oriented Evaluation of Lightweight IMU-Based Human Activity Recognition: On-Device Efficiency and Streaming Feasibility},
  journal = {Sensors},
  year = {2026},
  doi = {10.3390/s26175678}
}
Kostenlos starten →