The Physical Activity Assessment Using Wearable Sensors (PAAWS) Dataset: Labeled Laboratory and Free-living...

Area of Science:

• Human Activity Recognition
• Wearable Sensor Technology
• Digital Health

Background:

• Poor sleep and sedentary behaviors are linked to chronic diseases and reduced quality of life.
• Wearable sensors can monitor physical activity, sedentary behavior, and sleep in free-living conditions.
• Current human activity recognition algorithms struggle with real-world data due to lab-based training.

Purpose of the Study:

• To introduce the PAAWS R1 dataset, a multimodal, multi-sensor resource for human activity recognition.
• To provide a dataset enabling direct comparison of algorithms across different collection protocols and free-living conditions.
• To facilitate the development of robust algorithms for health research and mobile computing interventions.

Main Methods:

• Collected ~4 hours of semi-naturalistic activities from 252 individuals.
• Collected ~7 days of 24-hour free-living activities from 20 adults.
• Annotated waking activities with second-by-second video ground-truth labels and sleep stages from PSG data.

Main Results:

• The PAAWS R1 dataset includes diverse, real-world activity and sleep data.
• High-resolution annotations capture realistic activity transitions and sleep stages.
• The dataset supports direct algorithmic comparisons across varied collection settings.

Conclusions:

• The PAAWS dataset is a valuable resource for advancing human activity recognition.
• It enables the development of more robust algorithms for free-living conditions.
• This work supports improved health research and the creation of novel mobile health applications.