Related Experiment Video
Updated: Oct 4, 2025
08:45
Measuring the Kinematics of Daily Living Movements with Motion Capture Systems in Virtual Reality
Published on: April 5, 2018
7.8K
Data-Driven Classification of Human Movements in Virtual Reality-Based Serious Games: Preclinical Rehabilitation
Roni Barak Ventura1, Kora Stewart Hughes1, Oded Nov2
1Department of Mechanical and Aerospace Engineering, New York University Tandon School of Engineering, Brooklyn, NY, United States.
JMIR Serious Games
|February 10, 2022
Summary
This study introduces a novel virtual reality (VR) citizen science approach for telerehabilitation, accurately classifying bimanual movements to assess motor performance. This method enhances patient engagement and supports scientific advancement.
Area of Science:
- Rehabilitation Medicine
- Human-Computer Interaction
- Virtual Reality
Background:
- Sustained engagement is crucial for telerehabilitation success, but patient motivation and adherence are common challenges.
- Gamification of physical exercises is a strategy to improve engagement in telerehabilitation.
- A citizen science approach using virtual reality (VR) is proposed to enhance engagement by having patients contribute to scientific causes.
Purpose of the Study:
- To present a novel methodology for remote identification and classification of human movements for automatic motor performance assessment in telerehabilitation.
- To develop a data-driven approach within a citizen science software for bimanual training in VR.
- To enable users to contribute to citizen science projects while performing rehabilitation exercises in VR.
Main Methods:
- Nine healthy individuals used commercial VR gaming devices to interact with citizen science software.
- A calibration phase adapted software sensitivity to individual range of motion across three anatomical planes.
- Principal component analysis and a bagged trees ensemble classifier were used to analyze and classify movement data.
Main Results:
- The movement classification achieved high accuracy, reaching 99.9%.
- Elbow flexion was the most accurately classified movement at 99.2%.
- Horizontal shoulder abduction to the right was the most misclassified movement at 98.8% accuracy.
Conclusions:
- Coordinated bimanual movements in VR can be accurately classified.
- This approach provides a foundation for developing motion analysis algorithms for VR-mediated telerehabilitation.
- The citizen science model in VR shows promise for enhancing telerehabilitation engagement and effectiveness.
