EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Real-Time Human Activity Recognition on Edge Devices for Industrial Applications

Jamil Ahmad, Sergio Leggieri, Darwin Caldwell, Christian Di Natali

IEEE Sensors Journal · 15.05.2026 · 26 (10) · 15701-15715

DOI: 10.1109/jsen.2026.3678826 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Sensors Journal, erschienen am 15.05.2026. Verfasst von Jamil Ahmad und 3 weiteren Autor:innen. Die Arbeit wurde bisher 2-mal zitiert (Crossref). Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: BEEYONDERS project, European Union’s Horizon Europe research and innovation programme.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

The human activity recognition (HAR) is increasingly used to enhance exoskeleton control, collaborative robotics, and ergonomic assessments by providing high-level motion insights for industrial applications. However, deploying HAR on portable edge devices with limited computational resources presents challenges, highlighting the need for lightweight algorithms tailored to industry. This study proposes and compares low-powered lightweight HAR models capable of real-time (RT) operation on portable wearable systems. The models exploit sensor fusion from a full-body wearable system, the smart suit, designed for construction workers, with the goal of enabling accurate RT activity recognition while preserving the natural human motion frequency (NHMF) of 5 Hz. The impact of window size (WS) selection and the inclusion of transitional activities was analyzed to evaluate multiple machine learning (ML) [support vector machine (SVM), neural network (NN), and ensemble (ES)] and deep learning (DL) [convolutional recurrent NN (CRNN), bidirectional CRNN (BiCRNN), bidirectional long short-term memory (BiLSTM), and CNNTransformer] models combined with signal optimization techniques. Performance was assessed in terms of accuracy, F1-score, training time, inference time, FLOPs, energy consumption, and model size. A dedicated industrial dataset was iteratively constructed using a novel elementary–transition–elementary (ETE) recording technique, which simplified the recording and labeling of complex tasks and supports dataset scalability. The model generalization was evaluated using a leave-one-subject-out (LOSO) protocol, demonstrating that selected models achieved a maximum accuracy of 92.7% under realistic deployment constraints, enabling up to ten classifications per second.

Förderung

  • BEEYONDERS project
  • European Union’s Horizon Europe research and innovation programme

Zitieren

APA

Ahmad, J., Leggieri, S., Caldwell, D., & Natali, C. D. (2026). Real-Time Human Activity Recognition on Edge Devices for Industrial Applications. IEEE Sensors Journal, 26(10), 15701-15715. https://doi.org/10.1109/jsen.2026.3678826

Vancouver

Ahmad J, Leggieri S, Caldwell D, Natali CD. Real-Time Human Activity Recognition on Edge Devices for Industrial Applications. IEEE Sensors Journal. 2026;26(10):15701-15715. doi:10.1109/jsen.2026.3678826

BibTeX
@article{Ahmad2026,
  author = {Ahmad, Jamil and Leggieri, Sergio and Caldwell, Darwin and Natali, Christian Di},
  title = {Real-Time Human Activity Recognition on Edge Devices for Industrial Applications},
  journal = {IEEE Sensors Journal},
  year = {2026},
  volume = {26},
  number = {10},
  pages = {15701--15715},
  doi = {10.1109/jsen.2026.3678826}
}
Kostenlos starten →