Calibrationless monocular vision musculoskeletal simulation during gait.
1Department of Research and Development, ORGO, 2-7 Odori W18, Chuo-ku, Sapporo, 061-1136, Japan.
Heliyon
|June 13, 2024
Summary
This study validates using single-camera (monocular) video for markerless motion capture and musculoskeletal simulation. Results show comparable accuracy to multi-camera systems, enabling accessible gait analysis for clinicians.
Area of Science:
- Biomechanics
- Computer Vision
- Human Movement Analysis
Background:
- Markerless motion capture simplifies data acquisition.
- Previous methods like OpenCap used multiple devices.
- Monocular vision approaches offer potential for easier motion capture but require validation for musculoskeletal simulation.
Purpose of the Study:
- To validate musculoskeletal simulation accuracy using a monocular vision approach.
- To assess the feasibility of single-camera gait analysis.
- To compare monocular vision-based simulation outcomes with traditional motion capture.
Main Methods:
- Reconstructed 3D human motion from single videos using the SMPL model.
- Generated virtual marker data from the SMPL model.
- Performed inverse kinematics, GRF prediction, inverse dynamics, and static optimization.
- Calculated Mean Absolute Errors (MAE) against marker-based motion capture data.
Main Results:
- Mean Absolute Errors (MAE) were 8.4° for joint angles, 5.0% bodyweight for GRF, 1.1% bodyweight*height for joint moments, and 0.11 for muscle activations.
- Some outcomes were comparable to multi-camera systems (OpenCap).
- The monocular approach achieved larger, yet comparable, overall MAE.
Conclusions:
- Monocular vision-based motion capture and musculoskeletal simulation are valid for gait analysis.
- This approach requires no prior preparation, benefiting clinical applications.
- It offers a simplified, accessible method for quantifying daily gait assessments.
