EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Robust Feature Representation Using Multi-Task Learning for Human Activity Recognition

Behrooz Azadi iD, Michael Haslgrübler iD, Bernhard Anzengruber-Tanase, Georgios Sopidis iD, Alois Ferscha

Sensors · 21.01.2024 · 24 (2) · 681

DOI: 10.3390/s24020681 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Sensors, online erschienen am 21.01.2024. Verfasst von Behrooz Azadi und 4 weiteren Autor:innen. Beteiligte Einrichtung: Pro2Future GmbH, Altenberger Strasse 69, 4040 Linz, Austria. Hauptthemen laut MEDLINE: Learning, Skiing. Die Arbeit wurde bisher 23-mal zitiert (Crossref) und nennt 74 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: The Austrian Research Promotion Agency, Johannes Kepler Open Access Publishing Fund, FFG.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Learning underlying patterns from sensory data is crucial in the Human Activity Recognition (HAR) task to avoid poor generalization when coping with unseen data. A key solution to such an issue is representation learning, which becomes essential when input signals contain activities with similar patterns or when patterns generated by different subjects for the same activity vary. To address these issues, we seek a solution to increase generalization by learning the underlying factors of each sensor signal. We develop a novel multi-channel asymmetric auto-encoder to recreate input signals precisely and extract indicative unsupervised futures. Further, we investigate the role of various activation functions in signal reconstruction to ensure the model preserves the patterns of each activity in the output. Our main contribution is that we propose a multi-task learning model to enhance representation learning through shared layers between signal reconstruction and the HAR task to improve the robustness of the model in coping with users not included in the training phase. The proposed model learns shared features between different tasks that are indeed the underlying factors of each input signal. We validate our multi-task learning model using several publicly available HAR datasets, UCI-HAR, MHealth, PAMAP2, and USC-HAD, and an in-house alpine skiing dataset collected in the wild, where our model achieved 99%, 99%, 95%, 88%, and 92% accuracy. Our proposed method shows consistent performance and good generalization on all the datasets compared to the state of the art.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

LearningSkiingHumansCoping SkillsHuman ActivitiesRecognition, Psychology

Stichwörter der Autor:innen

alpine skiingdeep learninghuman activity recognitionmulti-task learningrepresentation learningwearable

Autor:innen und Institutionen

  • Pro2Future GmbH, Altenberger Strasse 69, 4040 Linz, Austria.
  • Institute of Pervasive Computing, Johannes Kepler University, Altenberger Straße 69, 4040 Linz, Austria.

ORCID-Profile: Behrooz Azadi, Michael Haslgrübler, Georgios Sopidis

Förderung

  • The Austrian Research Promotion Agency
  • Johannes Kepler Open Access Publishing Fund
  • FFG
  • Austrian Federal Ministry for Digital and Economic Affairs and of the Provinces of Upper Austria and Styria
  • Austrian Research Promotion Agency FFG

Zitieren

APA

Azadi, B., Haslgrübler, M., Anzengruber-Tanase, B., Sopidis, G., & Ferscha, A. (2024). Robust Feature Representation Using Multi-Task Learning for Human Activity Recognition. Sensors, 24(2), 681. https://doi.org/10.3390/s24020681

Vancouver

Azadi B, Haslgrübler M, Anzengruber-Tanase B, Sopidis G, Ferscha A. Robust Feature Representation Using Multi-Task Learning for Human Activity Recognition. Sensors. 2024;24(2):681. doi:10.3390/s24020681

BibTeX
@article{Azadi2024,
  author = {Azadi, Behrooz and Haslgrübler, Michael and Anzengruber-Tanase, Bernhard and Sopidis, Georgios and Ferscha, Alois},
  title = {Robust Feature Representation Using Multi-Task Learning for Human Activity Recognition},
  journal = {Sensors},
  year = {2024},
  volume = {24},
  number = {2},
  pages = {681},
  doi = {10.3390/s24020681}
}
Kostenlos starten →