Related Experiment Video
Updated: May 7, 2026
08:56
Clinical Assessment of Spatiotemporal Gait Parameters in Patients and Older Adults
Published on: November 7, 2014
14.0K
Gait stability prediction through synthetic time-series and vision-based data.
Mauricio C Cordeiro1, Ciaran O Cathain2,3,4, Vitor B Nascimento5
1Department of Engineering & Informatics, Technological University of the Shannon, Athlone, Ireland.
Frontiers in Sports and Active Living
|August 29, 2025
Summary
Synthetic data accurately replicates older adult gait parameters, enabling models trained on it to outperform real-world data models for predicting gait stability. Key predictors include step width, BMI, and fall history.
Area of Science:
- Biomechanics
- Computer Vision
- Gerontology
Background:
- Assessing gait stability in older adults is difficult due to limited data and complex measurements, especially in resource-limited settings.
- Existing methods often require specialized equipment, hindering widespread application among vulnerable populations.
Purpose of the Study:
- To investigate if synthetic data can accurately replicate gait parameters in older adults.
- To evaluate the effectiveness of models trained on synthetic data for predicting the Margin of Stability (MoS).
- To identify key biomechanical features influencing MoS predictions in older adults.
Main Methods:
- A constraint-based synthetic data generation (SDG) methodology was employed using a public dataset of healthy older adults.
- Gait analysis utilized smartphone-captured videos and the MediaPipe algorithm for landmark extraction, ensuring low-cost accessibility.
- The SDG approach preserved biomechanical relationships through metadata configuration and rank correlation constraints.
Main Results:
- The synthetic data achieved high fidelity (97.09%) in replicating gait parameters and maintaining biomechanical relationships.
- Models trained solely on synthetic data (TSTR) demonstrated superior performance over real data-trained models (TRTR), with significant reductions in prediction errors (RMSE, MAE, MSE) and improved variance explanation (R²).
- SHAP analysis confirmed that the synthetic data approach aligned feature attributions with established biomechanical principles, highlighting step width, BMI, and fall history as crucial predictors.
Conclusions:
- Synthetic data generation is a viable method to overcome data scarcity in gait analysis for older adults.
- Synthetic data-trained models significantly enhance the prediction accuracy of gait stability (MoS) compared to models trained on real data.
- This approach, leveraging accessible computer vision technology, offers a promising pathway for improved clinical gait assessment in resource-limited environments.
