Research Repository

Comparing shallow, deep, and transfer learning in predicting joint moments in running.

Liew, Bernard XW and Rügamer, David and Zhai, Xiaojun and Wang, Yucheng and Morris, Susan and Netto, Kevin (2021) 'Comparing shallow, deep, and transfer learning in predicting joint moments in running.' Journal of Biomechanics, 129. p. 110820. ISSN 0021-9290

[img] Text
paper_R1_v3.pdf - Accepted Version
Restricted to Repository staff only until 24 October 2022.
Available under License Creative Commons Attribution Non-commercial No Derivatives.

Download (181kB) | Request a copy

Abstract

Joint moments are commonly calculated in biomechanics research and provide an indirect measure of muscular behaviors and joint loads. However, joint moments cannot be easily quantified clinically or in the field, primarily due to challenges measuring ground reaction forces outside the laboratory. The present study aimed to compare the accuracy of three different machine learning (ML) techniques - functional regression [ ML<sub>fregress</sub> ], a deep neural network (DNN) built from scratch [ ML<sub>DNN</sub> ], and transfer learning [ ML<sub>TL</sub> ], in predicting joint moments during running. Data for this study came from an open-source dataset and two studies on running with and without external loads. Three-dimensional (3D) joint moments of the hip, knee, and ankle, were derived using inverse dynamics. 3D joint angle, velocity, and acceleration of the three joints served as predictors for each of the three ML techniques. Prediction performance was generally the best using ML<sub>DNN</sub>, and the worse using ML<sub>fregress</sub>. Absolute predictive performance was the best for sagittal plane moments, which ranged from a RMSE of 0.16 Nm/kg at the ankle using ML<sub>DNN</sub>, to a RMSE of 0.49Nm/kg at the knee using ML<sub>fregress</sub>. ML<sub>DNN</sub> resulted in the greatest improvement in relative prediction performance (relRMSE) by 20% compared to ML<sub>fregress</sub> for the ankle adduction-abduction moment. DNN with or without transfer learning was superior in predicting joint moments using kinematic inputs compared to functional regression. Synergizing ML with kinematic inputs has the potential to solve the constraints of obtaining high fidelity biomechanics data normally only possible during laboratory studies.

Item Type: Article
Uncontrolled Keywords: Running biomechanics; Inverse dynamics; Machine learning; Deep learning
Divisions: Faculty of Science and Health
Faculty of Science and Health > Computer Science and Electronic Engineering, School of
Faculty of Science and Health > Sport, Rehabilitation and Exercise Sciences, School of
SWORD Depositor: Elements
Depositing User: Elements
Date Deposited: 20 Dec 2021 14:40
Last Modified: 23 Sep 2022 19:48
URI: http://repository.essex.ac.uk/id/eprint/31909

Actions (login required)

View Item View Item