(Near) Real-time Data Analysis for Smart Sensor Shorts in Football

Master thesis (2023)

Authors

Q. Liu Electrical Engineering, Mathematics and Computer Science

Contributors

A. Bossche Electronic Instrumentation - EEMCS (mentor)
K.M.B. Jansen Emerging Materials - IDE (graduation committee member)
P.J. French Bio-Electronics - EEMCS (graduation committee member)
Faculty
Electrical Engineering, Mathematics and Computer Science, Electrical Engineering, Mathematics and Computer Science
Copyright: © 2023 Qingxin Liu
More Info
expand_more

Published Date

21-07-2023

Language

English

Reuse Rights

Other than for strictly personal use, it is not permitted to download, forward or distribute the text or part of it, without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license such as Creative Commons.

Faculty

Electrical Engineering, Mathematics and Computer Science

Abstract

Most injuries in football occur in the lower extremities due to high muscle stress. To prevent such injuries, the Dutch Football Association (KNVB) and the Delft University of Technology developed the Smart Sensor Shorts, an inertial sensor-based tracking system measuring the athlete’s lower body kinematics, to improve physical load estimates during training sessions and matches. However, the system currently only has offline data analysis software, which results in poor monitoring capability.

This thesis proposes a near real-time data analysis system for Smart Sensor Shorts, featuring an automatic sensor calibration module, a football-specific activity recognition module, and a user interface, to monitor users' lower limb movement and load during football training. The proposed automatic sensor-to-body calibration algorithm maintains a high calibration accuracy with an 18.92º(±5.74º) calibration error on average and simplifies the calibration process by leveraging detected standing and walking movements to estimate calibration parameters. The proposed gradient-boosting decision trees activity recognition model utilizes hip joint angles and joint angular velocities derived by the system to predict users' football-related activities, achieving an overall accuracy of 93.62%. The designed system processes the data recorded by IMUs in real time with a speed of 21 milliseconds per iteration and displays the calculated results related to the user's physical load on the user interface at a frame rate of 20 Hz.