ESTIMATION OF JOINT AND TENDON LOADING DURING RUNNING USING ARTIFICIAL IMU DATA AND UNSUPERVISED NEURAL NETWORKS

Authors

Sina David, Department of Human Movement Sciences, Vrije Universiteit AmsterdamFollow
Gabor Barton, Research Institute for Sport and Exercise Sciences, Liverpool John Moores UniversityFollow
Jasper Verheul, Cardiff School of Sport and Health Sciences, Cardiff Metropolitan UniversityFollow

Category

Wearable Technology

Document Type

Paper

Abstract

This study aimed to estimate lower-limb joint- and tendon loads during treadmill running by combining artificial IMU (artIMU) data of four virtually placed sensors on the shanks and feet with a self-organising neural network approach. To achieve this, we simulated IMU (artIMU) data from marker trajectories of 28 runners, running at 2.5, 3.5, and 4.5 m/s on a treadmill. A Kohonen self-organising map was trained with the artIMU data, and the joint and tendon loading was reconstructed as the hidden variables of the network. A leave-one-subject-out cross-validation resulted in a good to excellent estimation accuracy (R² > 0.87 and nRMSE

Recommended Citation

David, Sina; Barton, Gabor; and Verheul, Jasper (2024) "ESTIMATION OF JOINT AND TENDON LOADING DURING RUNNING USING ARTIFICIAL IMU DATA AND UNSUPERVISED NEURAL NETWORKS," ISBS Proceedings Archive: Vol. 42: Iss. 1, Article 255.
Available at: https://commons.nmu.edu/isbs/vol42/iss1/255