EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Activity Recognition from Daily-Life Sounds Using Unsupervised Learning with Dirichlet Multinomial Mixture Models

Ken Sadohara iD, Natsuki Miyata iD

Sensors · 2026

DOI: 10.3390/s26051509 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Sensors, 2026. Verfasst von Ken Sadohara und 1 weiteren Autor:innen. Beteiligte Einrichtung: National Institute of Advanced Industrial Science and Technology (AIST), 2-3-26 Aomi, Koto-ku, Tokyo 135-0064, Japan. Hauptthemen laut MEDLINE: Unsupervised Machine Learning, Sound, Activities of Daily Living, Pattern Recognition, Automated. Die Arbeit nennt 47 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: New Energy and Industrial Technology Development Organization.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

To support ambient assisted living for the elderly living alone, we investigate a method for recognizing daily activities from household sounds. To reduce the cost of building an activity-recognition model, we adopt an unsupervised learning approach based on a Dirichlet multinomial mixture model. The model represents the generative process of neural audio codec codes conditioned on latent activities. We further extend the model to handle multiple streams of codes corresponding to different sound directions. This extension enables the formation of more accurate activity clusters, partly because code occurrence patterns exhibit burstiness. The proposed approach is expected to serve as a key component for constructing an activity recognition system that requires minimal labeled data and a small number of user inquiries.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

Unsupervised Machine LearningSoundActivities of Daily LivingPattern Recognition, AutomatedHumansAlgorithms

Stichwörter der Autor:innen

Dirichlet multinomial mixture model (DMM)acoustic scene classificationactivity recognitionambient assisted livingbespoke systemburstinessneural audio codectopic model

Autor:innen und Institutionen

  • National Institute of Advanced Industrial Science and Technology (AIST), 2-3-26 Aomi, Koto-ku, Tokyo 135-0064, Japan.

ORCID-Profile: Ken Sadohara, Natsuki Miyata

Förderung

  • New Energy and Industrial Technology Development Organization

Zitieren

APA

Sadohara, K., & Miyata, N. (2026). Activity Recognition from Daily-Life Sounds Using Unsupervised Learning with Dirichlet Multinomial Mixture Models. Sensors. https://doi.org/10.3390/s26051509

Vancouver

Sadohara K, Miyata N. Activity Recognition from Daily-Life Sounds Using Unsupervised Learning with Dirichlet Multinomial Mixture Models. Sensors. 2026. doi:10.3390/s26051509

BibTeX
@article{Sadohara2026,
  author = {Sadohara, Ken and Miyata, Natsuki},
  title = {Activity Recognition from Daily-Life Sounds Using Unsupervised Learning with Dirichlet Multinomial Mixture Models},
  journal = {Sensors},
  year = {2026},
  doi = {10.3390/s26051509}
}
Kostenlos starten →