Related Experiment Video
Updated: Aug 21, 2025
10:52
Simulation of Human-induced Vibrations Based on the Characterized In-field Pedestrian Behavior
Published on: April 13, 2016
8.9K
Predicting vertical ground reaction forces from 3D accelerometry using reservoir computers leads to accurate gait
Margit M Bach1, Nadia Dominici1, Andreas Daffertshofer1
1Department of Human Movement Sciences, Faculty of Behavioural and Movement Sciences, Amsterdam Movement Sciences and Institute of Brain and Behaviour Amsterdam, Vrije Universiteit Amsterdam, Amsterdam, Netherlands.
Frontiers in Sports and Active Living
|November 17, 2022
Summary
This study uses machine learning with accelerometers to accurately detect gait events like foot-off during walking and running. This method combines low-cost sensors with AI for reliable locomotion analysis outside the lab.
Area of Science:
- Biomechanics
- Wearable Technology
- Machine Learning
Background:
- Accelerometers are accessible for outdoor gait analysis.
- Accurate detection of isolated gait events, particularly foot-off during running, remains a challenge.
Purpose of the Study:
- To develop and validate a machine learning approach for predicting vertical ground reaction forces from accelerometer data.
- To enable accurate detection of gait events using these predicted forces.
Main Methods:
- Collected shank accelerometer and ground reaction force data from 21 adults during walking and running.
- Trained a reservoir computer model on segmented gait data.
- Implemented a two-step process: ML prediction of forces, followed by force-based event detection.
Main Results:
- The reservoir computer accurately predicted vertical ground reaction forces from continuous gait data.
- Foot contact and foot-off event detection showed high accuracy compared to gold-standard methods.
- The model performed well despite being trained on limited data.
Conclusions:
- Combining accelerometry with machine learning offers a robust solution for detecting isolated gait events.
- This approach is effective across different modes of locomotion, including running and walking.
- The findings support the use of wearable sensors and AI for advanced gait analysis.
