Related Experiment Video
Updated: Apr 11, 2026
06:49
Evaluation of a Smartphone-based Human Activity Recognition System in a Daily Living Environment
Published on: December 11, 2015
9.4K
Validation and Reliability of a Classification Method to Measure the Time Spent Performing Different Activities
Marie-Ève Riou1, François Rioux2, Gilles Lamothe3
1Behavioural and Metabolic Research Unit (BMRU), School of Human Kinetics, Faculty of Health Sciences, University of Ottawa, Ottawa, Canada.
Plos One
|June 9, 2015
Summary
This study validates an activity classification model using accelerometers. The model accurately and reliably measures time spent in various activities across confined and unrestricted environments.
Area of Science:
- Biomedical Engineering
- Wearable Technology
- Activity Recognition
Background:
- Accurate quantification of physical activity is crucial for health research.
- Wearable sensors offer a promising method for objective activity monitoring.
- Validation of activity classification models in diverse environments is essential.
Purpose of the Study:
- To validate the performance and reliability of an activity classification model.
- To assess the model's accuracy in confined (CE) and unrestricted (UE) environments.
- To determine the model's effectiveness in measuring time spent on various activities.
Main Methods:
- Participants wore accelerometers (biaxial/triaxial) during pre-defined activities.
- A classification model was trained on Day 1 and validated on Day 2 (CE) and over 24 hours (UE).
- Model performance was evaluated against triaxial accelerometers using 6 or 8 features.
Main Results:
- Overall accuracy reached 94% in CE and 90% in UE.
- High sensitivity was observed for lying down (94%/95%), sitting (97%/89%), and walking (96%/78%).
- No significant performance difference was found between 6 and 8 features; results were highly reproducible.
Conclusions:
- The activity classification model demonstrates high accuracy and reproducibility.
- This approach has significant potential for quantifying physical activity in research settings.
- The validated model can be utilized in both controlled and real-world environments.
