Shank acceleration proxies for ground reaction force in athletic movements

Author: Andrew Vonow

Vonow, Andrew, 2020 Shank acceleration proxies for ground reaction force in athletic movements, Flinders University, College of Science and Engineering

Terms of Use: This electronic version is (or will be) made publicly available by Flinders University in accordance with its open access policy for student theses. Copyright in this thesis remains with the author. You may use this material for uses permitted under the Copyright Act 1968. If you are the owner of any included third party copyright material and/or you believe that any material has been made available without permission of the copyright owner please contact with the details.


Overuse injuries in elite netball have been reported at rates of up to 23.8 per 1000 hours of game play (Hume and Steele, 2000), with between 66 to 92% of these occurring at the lower limb (Hopper et al., 1995, McManus et al., 2006). Since acceleration and force are linearly dependent, this thesis aimed to determine whether a wearable accelerometer at the lower limb could be used to determine an objective level of impact dose. If so, and if dose could be quantified in terms of risk, the prediction could be used to monitor and prevent overuse injuries. There were 16 articles identified as having previously studied the relationship between ground reaction force and acceleration. For the 15 locations across the body that these articles used to predict impact force, five different kinds of proxy models had been considered: linear, logarithmic, Fourier, machine learning, and an unknown application. Variables that were most notably identified to affect prediction accuracy were technique, movement, and proxy type. These models were reported to have produced several very large (R^2>0.7) correlations between force and acceleration. Three of the literature-identified models were validated against a trial dataset collected in a previous generation of this project. Linear and logarithmic models were used to predict force events from acceleration, and a kind of machine learning was used to predict the entire waveform. The models that considered events correlated 14 different triaxial acceleration waveform events against 6 different force waveform events. The investigation considered accelerations from the thigh, shank, and ankle, with one accelerometer at each position on both legs. The strongest correlations were almost perfect (R^2>0.90) and were between the integration of vertical force, which is vertical impulse, and the integration of axial acceleration at the ankle. Although the shank and thigh also produced, on average, at least very large correlations, the ankle was deemed the optimal position for relating these waveforms. There were no events across the linear and logarithmic models that produced correlations that were consistently above R^2=0.36 (moderate). However, the machine learning model predicted the entire waveform with an accuracy of R^2=0.41, which was considered more favourable than the event-based linear and logarithmic models. As such, the investigation did not reproduce the literature-based results. If it can be shown in the future that impulse can indeed be used as an indicator of impact dose, and if these results can be reproduced, then lower limb acceleration may indeed be used as a proxy for monitoring impact dose and in overuse injury management and prevention.

Keywords: acceleration, force, ground reaction, proxy, surrogate, algorithm, approximation, athletic, movement, shank, lower limb, ankle, thigh, netball, overuse, injury, risk, impact dose, impulse, integration, logarithmic, linear, time series, filtering, review, analysis, wearable, device, inertial measurement unit

Subject: Engineering thesis

Thesis type: Masters
Completed: 2020
School: College of Science and Engineering
Supervisor: Dr. David Hobbs