Related Experiment Video
Updated: May 4, 2026
Physical Activity Measurement in Children Accepting Table Tennis Training
Published on: July 27, 2022
Fully automated waist-worn accelerometer algorithm for detecting children's sleep-period time separate from 24-h
Catrine Tudor-Locke1, Tiago V Barreira, John M Schuna
1Pennington Biomedical Research Center, 6400 Perkins Road, Baton Rouge, LA 70808, USA.
Insights
Accurately identifying sleep-period time in children using waist-worn accelerometers is crucial for physical activity analysis. A new fully automated algorithm (Algorithm 3) precisely estimates sleep time, outperforming previous methods.
Area of Science:
- Biomedical Engineering
- Pediatric Health Monitoring
- Activity Recognition
Background:
- Accurate sleep-period time identification is essential for analyzing 24-hour accelerometer data in children.
- Existing automated algorithms often require manual input or overestimate sleep duration.
Purpose of the Study:
- To evaluate the validity of a published automated algorithm for sleep-period time detection in children.
- To validate a refined, fully automated algorithm for improved sleep and activity data separation.
Main Methods:
- Compared expert visual analysis of waist-worn accelerometer data with three automated algorithms in 30 fourth-grade schoolchildren.
- Algorithm 1: Published algorithm; Algorithm 2: Included inclinometer data; Algorithm 3: Focused on automated bedtime/wake time detection.
Main Results:
- Algorithms 1 and 2 significantly overestimated sleep time (43 min and 90 min, respectively) compared to the criterion.
- Algorithm 3 demonstrated the smallest mean difference (2 min) and was not significantly different from expert visual inspection.
Conclusions:
- The fully automated Algorithm 3 provides a precise and valid method for estimating sleep-period time in children using accelerometer data.
- This algorithm will enhance the accurate differentiation of sleep from physical activity and sedentary behavior in research.
Abstract:
Analysis of 24-h waist-worn accelerometer data for physical activity and sedentary behavior requires that sleep-period time (from sleep onset to the end of sleep, including all sleep epochs and wakefulness after onset) is first identified. To identify sleep-period time in children in this study, we evaluated the validity of a published automated algorithm that requires nonaccelerometer bed- and wake-time inputs, relative to a criterion expert visual analysis of minute-by-minute waist-worn accelerometer data, and validated a refined fully automated algorithm. Thirty grade 4 schoolchildren (50% girls) provided 24-h waist-worn accelerometry data. Expert visual inspection (criterion), a published algorithm (Algorithm 1), and 2 additional automated refinements (Algorithm 2, which draws on the instrument's inclinometer function, and Algorithm 3, which focuses on bedtime and wake time points) were applied to a standardized 24-h time block. Paired t tests were used to evaluate differences in mean sleep time (expert criterion minus algorithm estimate). Compared with the criterion, Algorithm 1 and Algorithm 2 significantly overestimated sleep time by 43 min and 90 min, respectively. Algorithm 3 produced the smallest mean difference (2 min), and was not significantly different from the criterion. Relative to expert visual inspection, our automated Algorithm 3 produced an estimate that was precise and within expected values for similarly aged children. This fully automated algorithm for 24-h waist-worn accelerometer data will facilitate the separation of sleep time from sedentary behavior and physical activity of all intensities during the remainder of the day.
