Related Experiment Video
Updated: Jan 2, 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
8.6K
Optimization and Validation of an Adjustable Activity Classification Algorithm for Assessment of Physical Behavior in
Wouter Bijnens1, Jos Aarts1, An Stevens1
1Instrument Development, Engineering and Evaluation, Maastricht University, PO Box 616, 6200 MD Maastricht, The Netherlands.
Sensors (Basel, Switzerland)
|December 11, 2019
Summary
This study presents an adjustable physical activity classification algorithm for activity trackers. The algorithm accurately classifies sedentary and standing behaviors, aiding researchers and clinicians in selecting appropriate devices.
Area of Science:
- Biomedical Engineering
- Wearable Technology
- Human Movement Analysis
Background:
- Lack of transparency in algorithm and validation methods hinders selection of physical activity trackers.
- Need for adaptable algorithms to suit diverse populations and tracker placements.
Purpose of the Study:
- To transparently present an adjustable physical activity classification algorithm.
- To discriminate between dynamic, standing, and sedentary behaviors.
- To optimize and validate the algorithm for elderly populations wearing trackers on the upper leg under simulated free-living conditions.
Main Methods:
- Developed an adjustable algorithm with parameters for data segmentation window size and physical activity threshold.
- Validated the algorithm using a fixed activity protocol (FAP) and a simulated free-living protocol (SFP).
- Assessed performance using percentage error (PE) and absolute percentage error (APE) with 20 participants for FAP and 20 for SFP.
Main Results:
- Standing and sedentary behaviors classified within acceptable limits (±10% error) in both fixed and simulated free-living conditions.
- Dynamic behavior classification was within acceptable limits under fixed conditions.
- Dynamic behavior classification showed limitations under simulated free-living conditions.
Conclusions:
- The adjustable algorithm offers transparency and facilitates activity tracker selection for researchers and clinicians.
- The proposed approach can accelerate the development of new applications for physical activity monitoring.
- Further refinement is needed for dynamic behavior classification in simulated free-living environments.
