Home // HEALTHINFO 2023, The Eighth International Conference on Informatics and Assistive Technologies for Health-Care, Medical Support and Wellbeing // View article

Attempt for Estimation of Vertical Ground Reaction Force by Deep Learning with Time Factor from 2D Walking Images

Authors:
Takeshi Mochizuki
Kyoko Shibata

Keywords: Gait Analysis; Ground Reaction Force; Estimation; 3D CNN; Single Camera.

Abstract:
Ground reaction force data are useful for evaluating gait stability, but only specialized institutions can measure it because installed force plates are often used to measure it with high accuracy. Therefore, this report proposes an easy method for estimating ground reaction forces using images captured by a widely available device. In a previous report, we created an algorithm to estimate the ground reaction force from images using 2D Convolutional Neural Network (CNN), one of the deep learning. The results showed that if a deep learning model is created in advance, the estimation of vertical ground reaction forces can have an 8% to 14% error to body weight. To further improve accuracy, this report creates training data that include the time factor and performs vertical ground reaction force estimation by 3D CNN. The training data used in this report, the voxel data were created using images at the time of estimation and images prior to that time to incorporate the time factor. The results, estimation of ground reaction force resulted in a 15% error relative to body weight and did not improve accuracy. Since overlearning occurred in all deep learning models, we suppose that accuracy was not improved due to insufficient training data or bias.

Pages: 22 to 25

Copyright: Copyright (c) IARIA, 2023

Publication date: November 13, 2023

Published in: conference

ISSN: 2519-8491

ISBN: 978-1-68558-105-3

Location: Valencia, Spain

Dates: from November 13, 2023 to November 17, 2023