Towards a Markerless 3D Pose Estimation Tool

Published

April 28, 2023

Author(s)

Amaan Rahman, Mili Shah, Ya-Shian Li-Baboud, Ann Virts

Abstract

Evaluation of exoskeleton performance benefts from standards to verify proper functionality and safety. Currently, there are limited evaluation methods for exoskeletons. Measurement methods to evaluate human-exoskeleton kinematics include optical tracking systems (OTS) and inertial measurement units (IMUs). However, OTS and IMUs can be intrusive, requiring the attachment of markers or sensors. This research focuses on investigating markerless 3D pose estimation algorithms with low-cost red, green, blue (RGB) cameras to determine their viability as methods for tracking human joint positions and deriving skeletal frame orientations.We present a tool that utilizes state-of-the-art 3D pose estimation algorithms to generate 3D pose estimation data. Future experiments will be performed to evaluate the viability of 3D pose estimation algorithms as markerless methods for joint position and orientation estimation.

Conference Dates

April 23-28, 2023

Conference Location

Hamburg, DE

Conference Title

Human-Computer Interaction (HCI)

Pub Type

Conferences

Download Paper

https://doi.org/10.1145/3544549.3583950

Local Download

Keywords

pose estimation, exoskeleton evaluation, kinematics, temporal convolutional networks

Citation

Rahman, A. , Shah, M. , Li-Baboud, Y. and Virts, A. (2023), Towards a Markerless 3D Pose Estimation Tool, Human-Computer Interaction (HCI), Hamburg, DE, [online], https://doi.org/10.1145/3544549.3583950, https://tsapps.nist.gov/publication/get_pdf.cfm?pub_id=934902 (Accessed December 12, 2024)