Related Experiment Video
Updated: Mar 14, 2026
07:24
Home-Based Monitor for Gait and Activity Analysis
Published on: August 8, 2019
7.4K
Continuous assessment of daily-living gait using self-supervised learning of wrist-worn accelerometer data
Yonatan E Brand1,2, Aron S Buchman3, Felix Kluge4
1School of Biomedical Engineering, Tel Aviv University, Tel Aviv, Israel.
NPJ Digital Medicine
|March 13, 2026
Summary
ElderNet, a deep-learning model, now estimates gait metrics like speed and stride length from wrist accelerometers. This technology offers a scalable solution for assessing mobility in older adults and patients with gait impairments.
Area of Science:
- Gerontology
- Biomedical Engineering
- Machine Learning
Background:
- Physical activity and mobility are crucial for healthy aging and diverse health outcomes.
- Wrist-worn accelerometers are common for monitoring physical activity, but gait metric estimation from this data is difficult.
- Existing methods struggle to accurately assess gait parameters using wearable sensors.
Purpose of the Study:
- To extend the ElderNet deep-learning model for estimating gait metrics from wrist accelerometry data.
- To validate the performance of ElderNet in diverse populations, including older adults and individuals with gait impairments.
- To compare ElderNet's accuracy against state-of-the-art methods and lower-back sensor models.
Main Methods:
- Utilized ElderNet, a self-supervised deep-learning model, to analyze wrist accelerometry data.
- Validated the model on datasets from 819 older adults (Rush-Memory-and-Aging-Project) and 85 individuals with gait impairments (Mobilise-D) across six medical centers.
- Compared ElderNet's performance against existing gait analysis techniques and models using different sensor placements.
Main Results:
- ElderNet achieved high accuracy in gait speed estimation (absolute error: 8.82 cm/s, ICC: 0.87) in the Mobilise-D cohort, outperforming other methods.
- The model demonstrated superior performance in estimating cadence and stride length compared to competing approaches.
- ElderNet effectively classified mobility disability (AUC = 0.80), surpassing conventional gait and physical activity metrics.
Conclusions:
- ElderNet shows significant potential as a scalable tool for gait assessment using readily available wrist-worn devices.
- The model offers a promising approach for monitoring mobility in aging populations and individuals with clinical gait impairments.
- This advancement facilitates more accessible and accurate gait analysis outside of laboratory settings.
