Aggregation Periods Influence Step Count Error in Low-Power Wearables.

Area of Science:

• Wearable technology
• Biomedical engineering
• Human activity recognition

Background:

• Low-power wearables are crucial for long-term physical activity monitoring.
• Data aggregation in low-power devices can compromise measurement accuracy.
• Optimizing sensor settings is key for reliable free-living data.

Purpose of the Study:

• To evaluate a new low-power wearable (LPW) for step monitoring.
• To compare LPW performance against a research-grade sensor (RGS).
• To assess the impact of different aggregation periods (APs) on step count accuracy.

Main Methods:

• Thirty-two participants wore both LPW and RGS devices.
• LPW data were collected using 10 min, 1 min, and 10 s APs.
• Wear time detection, total daily step error, and granular data accuracy were analyzed.

Main Results:

• High sensitivity (0.96) and specificity (0.98) for wear time detection across all APs.
• No significant difference in total daily step count error between APs.
• The 10 min AP showed greater undercounting and variability, especially for fragmented activity bouts.

Conclusions:

• Aggregation period significantly impacts the accuracy of granular step count data, not total daily counts.
• Longer APs may obscure short activity bursts, leading to underestimation.
• Careful selection of APs is essential for accurate low-power wearable data in real-world settings.