Related Experiment Video
Updated: Aug 31, 2025
06:21
Postural Organization of Gait Initiation for Biomechanical Analysis Using Force Platform Recordings
Published on: July 26, 2022
2.6K
Comparing the accuracy of open-source pose estimation methods for measuring gait kinematics.
Edward P Washabaugh1, Thanikai Adhithiyan Shanmugam2, Rajiv Ranganathan3
1Department of Biomedical Engineering, Wayne State University, Detroit, MI, USA; Department of Physical Medicine and Rehabilitation, Michigan Medicine, Ann Arbor, MI, USA.
Gait & Posture
|August 21, 2022
Summary
Open-source pose estimation software like OpenPose offers a cost-effective alternative for gait analysis. It accurately measures kinematics and spatiotemporal parameters, making it valuable for clinical and research settings.
Area of Science:
- Biomechanics
- Computer Vision
- Machine Learning
Background:
- Marker-based motion capture for gait analysis is costly and complex.
- Open-source pose estimation offers a potential solution to reduce costs and improve accessibility.
- Machine learning in pose estimation minimizes the need for specialized equipment.
Purpose of the Study:
- To compare the accuracy of open-source pose estimation software for gait analysis.
- To evaluate the measurement of kinematics and spatiotemporal gait parameters.
- To determine the most suitable open-source method for clinical and research applications.
Main Methods:
- Utilized an existing dataset of 32 participants with video and motion capture data.
- Analyzed sagittal plane videos using OpenPose, Tensorflow MoveNet Lightning, Thunder, and DeepLabCut.
- Calculated hip/knee kinematics and spatiotemporal parameters, comparing errors against marker-based systems.
Main Results:
- Pose estimation methods showed varying accuracy in kinematic measurements.
- OpenPose and MoveNet Thunder demonstrated the highest accuracy for hip kinematics.
- OpenPose was most accurate for knee kinematics, with low errors in spatiotemporal parameters.
Conclusions:
- OpenPose significantly outperforms other tested platforms for gait analysis.
- Open-source pose estimation, particularly OpenPose, can serve as a viable alternative to conventional motion capture.
- This technology can enhance gait analysis in clinical and research settings where traditional systems are unavailable.
