Integrating 2D and 3D Human Pose Estimation Using Depth Camera in Cardio Exercises

Estimating the human keypoints, i.e. wrist, elbow, and shoulder in a precise manner is crucial when performing biomechanical analysis. The common method used in estimating the keypoints is 2D human pose estimation. The use of HPE can estimate the human keypoints from an image or videos. However, the...

Full description

Saved in:
Bibliographic Details
Published in2024 International Conference on Green Energy, Computing and Sustainable Technology (GECOST) pp. 255 - 259
Main Authors Wei Tiong, Alan Ka, Lim, King Hann, Sien Phang, Jonathan Then, Pang, Po Ken
Format Conference Proceeding
LanguageEnglish
Published IEEE 17.01.2024
Subjects
2D-to-3D Lifting
Biological system modeling
Biomechanics
Cameras
Deep Learning
Depth Camera
Human Pose Estimation
Pose estimation
Three-dimensional displays
Tracking
Wrist
Online AccessGet full text

Cover

More Information
Summary:Estimating the human keypoints, i.e. wrist, elbow, and shoulder in a precise manner is crucial when performing biomechanical analysis. The common method used in estimating the keypoints is 2D human pose estimation. The use of HPE can estimate the human keypoints from an image or videos. However, there are some challenges using 2D HPE approach. The challenges can lead to inaccurate estimating of human keypoints. The use of depth sensing camera able to provide the 3D information of the respective 2D keypoints. The combination of 2D information obtained from the HPE model and the 3D information of the respective key joints obtained from the depth sensing camera can effectively solve the issues encountered by 2D HPE. The use of human detection model and human pose estimation model are performed to detect key joints using deep learning approach. The detected key joints can subsequently be processed to study human biomechanical motion and analysis.
DOI:10.1109/GECOST60902.2024.10474635