Key result
A machine learning algorithm for a single thigh-mounted accelerometer achieved a 64% weighted average F-measure for classifying 11 physical behaviors in a simulated free-living validation.
Why the study?
Physical activity is emerging as an outcome measure, and newer analytical recognition methods using accelerometers can increase the degree of detail when monitoring physical behavior.
A machine learning algorithm using a single thigh-mounted accelerometer can accurately classify 5 physical behaviors, though performance drops when attempting to classify 11 behaviors.
Enhances PA classification granularity with thigh accelerometers; leaves open clinical validation in cardiovascular cohorts.
BACKGROUND Physical activity (PA) is emerging as an outcome measure. Accelerometers have become an important tool in monitoring physical behavior and newer analytical approaches of recognition methods increase the degree of details. OBJECTIVE The purpose of this study was to develop and validate an algorithm for classifying several daily physical behaviors using a single thigh-mounted accelerometer and a supervised machine learning scheme. METHODS We collected training data for adding further behavior classes to an existing algorithm. Combining data, we were potentially able to classify 11 behaviors, using a Random Forest learning scheme. We validated the algorithm through a simulated free-living procedure using chest-mounted cameras for establishing the ground truth. Furthermore, we adjusted our algorithm and compared the performance with a validated algorithm. RESULTS In the simulated free-living validation, the performance of the algorithm decreased to 64% as a weighted average for the 11 classes (F-measure). After reducing to 5 classes corresponding with the validated algorithm, the result revealed high performance in comparison with both the ground truth and the validated algorithm. CONCLUSIONS We developed an algorithm to classify 11 physical behaviors. We obtained high classification levels within specific behaviors, while others yielded lower classification potential.
No takes yet. Share an insight, caveat, or question.
Skovbjerg et al. (2022) studied this question. Machine learning-based algorithm for thigh-mounted accelerometers vs. Ground truth (chest-mounted cameras) and a validated algorithm was evaluated on Classification performance of 11 physical behaviors (F-measure). A machine learning algorithm for a single thigh-mounted accelerometer achieved a 64% weighted average F-measure for classifying 11 physical behaviors in a simulated free-living validation.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: