Related Experiment Video
Updated: Mar 17, 2026
11:21
Methodology for Establishing a Community-Wide Life Laboratory for Capturing Unobtrusive and Continuous Remote Activity and Health Data
Published on: July 27, 2018
8.9K
Unsupervised detection and analysis of changes in everyday physical activity data.
Gina Sprint1, Diane J Cook1, Maureen Schmitter-Edgecombe2
1School of Electrical Engineering and Computer Science, Washington State University, Pullman, WA, United States.
Journal of Biomedical Informatics
|July 30, 2016
Summary
This study introduces a Physical Activity Change Detection (PACD) framework to monitor lifestyle changes using wearable sensor data. PACD effectively identifies and analyzes physical activity shifts, aiding health goal tracking.
Area of Science:
- Biomedical Engineering
- Data Science
- Human-Computer Interaction
Background:
- Wearable sensors are increasingly used to monitor physical health and lifestyle changes.
- Automated detection of behavior changes from sensor data offers valuable monitoring and motivation.
- Existing methods lack a comprehensive framework for unsupervised physical activity change detection.
Purpose of the Study:
- To formalize the problem of unsupervised physical activity change detection.
- To introduce and evaluate the Physical Activity Change Detection (PACD) framework.
- To compare the effectiveness of different change detection algorithms within the PACD framework.
Main Methods:
- Developed the Physical Activity Change Detection (PACD) framework.
- Implemented algorithms for detecting, determining significance, and analyzing changes in time series data.
- Evaluated PACD using synthetic data and real-world Fitbit data from older adults.
Main Results:
- PACD successfully detected multiple changes in both synthetic and real-world datasets.
- The framework demonstrated the utility of proposed and existing change detection algorithms.
- Analysis revealed the effectiveness of PACD in tracking physical activity patterns.
Conclusions:
- The PACD framework provides a valuable tool for unsupervised, window-based change detection in physical activity data.
- The proposed algorithms and analysis methods can help track user activity and motivate health goal achievement.
- This approach has significant potential for health intervention studies and personal health monitoring.
