EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Real-Time Sensor-Based Human Activity Recognition for eFitness and eHealth Platforms

Łukasz Czekaj iD, Mateusz Kowalewski iD, Jakub Domaszewicz iD, Robert Kitłowski iD, Mariusz Szwoch iD, Włodzisław Duch iD

Sensors · 2024

DOI: 10.3390/s24123891 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Sensors, 2024. Verfasst von Łukasz Czekaj und 5 weiteren Autor:innen. Beteiligte Einrichtung: Aidmed, 80-254 Gdańsk, Poland. Hauptthemen laut MEDLINE: Telemedicine, Human Activities, Neural Networks, Computer. Die Arbeit wurde bisher 15-mal zitiert (Crossref) und nennt 29 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: National Centre for Research and Development (NCBiR), Poland, and the Polish National Science Center, National Centre for Research and Development (NCBiR), Poland, Polish National Science Center.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Human Activity Recognition (HAR) plays an important role in the automation of various tasks related to activity tracking in such areas as healthcare and eldercare (telerehabilitation, telemonitoring), security, ergonomics, entertainment (fitness, sports promotion, human–computer interaction, video games), and intelligent environments. This paper tackles the problem of real-time recognition and repetition counting of 12 types of exercises performed during athletic workouts. Our approach is based on the deep neural network model fed by the signal from a 9-axis motion sensor (IMU) placed on the chest. The model can be run on mobile platforms (iOS, Android). We discuss design requirements for the system and their impact on data collection protocols. We present architecture based on an encoder pretrained with contrastive learning. Compared to end-to-end training, the presented approach significantly improves the developed model’s quality in terms of accuracy (F1 score, MAPE) and robustness (false-positive rate) during background activity. We make the AIDLAB-HAR dataset publicly available to encourage further research.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

TelemedicineHuman ActivitiesNeural Networks, ComputerHumansExerciseAlgorithms

Stichwörter der Autor:innen

contrastive learningdeep networkshuman activity recognitionhuman–computer interactioninertial measurement unitmobile

Autor:innen und Institutionen

  • Aidmed, 80-254 Gdańsk, Poland.
  • Faculty of Electronics, Telecommunications and Informatics, Gdańsk University of Technology, 80-233 Gdańsk, Poland.
  • Department of Informatics, Institute of Engineering and Technology, Faculty of Physics, Astronomy & Informatics, Nicolaus Copernicus University, 87-100 Torun, Poland.

ORCID-Profile: Łukasz Czekaj, Mateusz Kowalewski, Jakub Domaszewicz, Robert Kitłowski, Mariusz Szwoch, Włodzisław Duch

Förderung

  • National Centre for Research and Development (NCBiR), Poland, and the Polish National Science Center
  • National Centre for Research and Development (NCBiR), Poland
  • Polish National Science Center

Zitieren

APA

Czekaj, Ł., Kowalewski, M., Domaszewicz, J., Kitłowski, R., Szwoch, M., & Duch, W. (2024). Real-Time Sensor-Based Human Activity Recognition for eFitness and eHealth Platforms. Sensors. https://doi.org/10.3390/s24123891

Vancouver

Czekaj Ł, Kowalewski M, Domaszewicz J, Kitłowski R, Szwoch M, Duch W. Real-Time Sensor-Based Human Activity Recognition for eFitness and eHealth Platforms. Sensors. 2024. doi:10.3390/s24123891

BibTeX
@article{Czekaj2024,
  author = {Czekaj, Łukasz and Kowalewski, Mateusz and Domaszewicz, Jakub and Kitłowski, Robert and Szwoch, Mariusz and Duch, Włodzisław},
  title = {Real-Time Sensor-Based Human Activity Recognition for eFitness and eHealth Platforms},
  journal = {Sensors},
  year = {2024},
  doi = {10.3390/s24123891}
}
Kostenlos starten →