Related Experiment Video
Updated: Nov 30, 2025
02:15
Predictive Measurement for Windlass Change in Length and Selected Treatment Outcomes in Chronic Plantar Fasciitis
Published on: March 1, 2024
740
Random forest algorithms for recognizing daily life activities using plantar pressure information: a smart-shoe study
Dian Ren1,2, Nathanael Aubert-Kato3,4, Emi Anzai5
1Department of Human and Environmental Sciences, Ochanomizu University, Tokyo, Japan.
Peerj
|November 16, 2020
Summary
Smart shoes using plantar pressure sensors can accurately recognize daily activities, achieving 89% success. Minimal sensor configurations maintain high performance, paving the way for advanced physical activity monitoring.
Area of Science:
- Biomedical Engineering
- Human-Computer Interaction
- Sports Science
Background:
- Wearable activity trackers offer new avenues for health promotion interventions.
- Smart clothing with multi-sensing capabilities presents novel opportunities for activity prediction.
- Accurate physical behavior evaluation relies on algorithms processing diverse smart device data.
Purpose of the Study:
- Develop an activity recognition algorithm using plantar pressure data from smart-shoe prototypes.
- Determine optimal hardware and software configurations for smart-shoe activity recognition.
Main Methods:
- Seventeen subjects performed nine distinct activities while wearing smart-shoe prototypes with seven pressure sensors.
- Plantar pressure data were collected, windowed, and analyzed using 167 features with a random forest model.
- Algorithm performance was evaluated by modulating window lengths, sensor configurations, and feature numbers.
Main Results:
- An optimal window length of 20 seconds yielded an 89% average success rate for activity recognition.
- Running activity achieved 100% sensitivity, while walking up a slope showed the lowest performance (63%).
- Reduced sensor (2-3) and feature (20) configurations maintained high performance (87% success).
Conclusions:
- High-performance human behavior recognition is achievable using solely plantar pressure data.
- Smart-shoe devices show potential for evaluating daily physical activities.
- Minimalist smart-shoe configurations with fewer sensors and features can ensure satisfactory performance.
