Optimizing Trajectories and Inverse Kinematics for Biomechanical Analysis of Markerless Motion Capture Data

Area of Science:

• Biomechanics
• Computer Vision
• Rehabilitation Science

Background:

• Markerless motion capture using computer vision and human pose estimation (HPE) promises wider access to precise movement analysis.
• Accurate tracking of outcomes and sensitive research tools are crucial for rehabilitation.
• Significant gaps exist in guiding design decisions for video-to-biomechanical data pipelines.

Purpose of the Study:

• To analyze critical steps in markerless motion capture pipelines for biomechanical analysis.
• To identify key features for accurate movement estimation in rehabilitation settings.
• To develop a user-friendly pipeline for obtaining precise biomechanical movement data.

Main Methods:

• Utilized recent human pose estimation (HPE) algorithms for dense keypoint detection.
• Employed implicit representations for smooth, anatomically constrained trajectory reconstruction.
• Implemented iterative optimization of biomechanical models and regularization for inverse kinematics (IK).

Main Results:

• A dense set of biomechanically-motivated keypoints improves accuracy.
• Implicit trajectory reconstruction enhances anatomical constraint for IK.
• Iterative model optimization and appropriate regularization are vital for precise biomechanical estimates.

Conclusions:

• The developed pipeline simplifies the extraction of accurate biomechanical movement estimates.
• This approach can significantly advance rehabilitation research and clinical practice through precise movement analysis.
• Optimizing HPE algorithms, trajectory reconstruction, and IK processes is key to reliable markerless motion capture.