Improving Gait Analysis Techniques with Markerless Pose Estimation Based on Smartphone Location

Area of Science:

• Biomechanics and Movement Science
• Computer Vision and Machine Learning

Background:

• Marker-based 3D motion capture systems are accurate for gait analysis but expensive and inaccessible.
• Markerless pose estimation presents a convenient, cost-effective alternative, though accuracy challenges persist.
• Limited research exists on camera location and orientation's impact on pose estimation accuracy for gait analysis.

Purpose of the Study:

• To compare the data collection accuracy of markerless pose estimation with marker-based systems for gait analysis.
• To assess how camera location and orientation influence the accuracy of gait assessment using pose estimation.
• To identify optimal camera placements for reliable gait analysis with markerless systems.

Main Methods:

• Utilized five smartphones for markerless pose estimation during gait analysis.
• Investigated the effect of varying camera locations and orientations on data collection accuracy.
• Compared joint angle data from pose estimation with a gold-standard marker-based system.

Main Results:

• Pose estimation showed differences below 5° in joint angles compared to marker-based systems, indicating acceptable accuracy.
• A strong correlation was found between markerless pose estimation and marker-based datasets, validating the approach.
• Optimal accuracy for hip and knee angles was achieved with cameras positioned front-diagonally; ankle angles were best from the lateral side.

Conclusions:

• Markerless pose estimation is an effective and accurate method for gait analysis, comparable to traditional marker-based systems.
• Strategic camera placement significantly impacts the quantitative accuracy of gait analysis using pose estimation.
• This study provides essential guidance for optimizing camera setups to enhance the reliability of markerless gait analysis.