Thigh-worn accelerometry: a comparative study of two no-code classification methods for identifying physical activity...

Area of Science:

• Biomedical Engineering
• Human Movement Science
• Wearable Technology

Background:

• Accurate assessment of free-living physical behavior is crucial for understanding health and well-being.
• Thigh-worn accelerometry offers a promising method for identifying activity types and postures.
• User-friendly, no-code software solutions are needed to increase the adoption of accelerometry.

Purpose of the Study:

• To evaluate the classification accuracy of two novel no-code software methods: SENS motion and ActiPASS.
• To compare the performance of these methods in identifying physical activities and postures.
• To assess their utility in both laboratory and free-living conditions.

Main Methods:

• 38 healthy adults wore thigh-mounted SENS motion accelerometers (12.5 Hz).
• Participants performed standardized laboratory activities and unrestricted free-living activities.
• Video recordings with chest-mounted cameras served as reference annotations for free-living data.
• Classification outputs from SENS motion and ActiPASS software were compared to reference labels.

Main Results:

• Analysis of 63.6 hours of activity data showed high agreement between algorithms and references.
• Cohen's kappa coefficients in free-living conditions were 0.86 for SENS motion and 0.92 for ActiPASS.
• Mean balanced accuracy ranged from 0.81 (cycling) to 0.99 (running) for SENS motion and 0.92 (walking) to 0.99 (sedentary) for ActiPASS.

Conclusions:

• Both SENS motion and ActiPASS no-code methods accurately classify basic physical activity types and postures.
• The methods demonstrate high accuracy even with relatively low sampling frequency data.
• Performance differences were noted, particularly in free-living cycling (SENS) and slow walking (ActiPASS), potentially due to differing activity class definitions.