No Cover Image

Journal article 64 views 24 downloads

Determining jumping performance from a single body-worn accelerometer using machine learning

Mark White, Neil Bezodis Orcid Logo, Jonathon Neville, Huw Summers Orcid Logo, Paul Rees Orcid Logo

PLOS ONE, Volume: 17, Issue: 2, Start page: e0263846

Swansea University Authors: Mark White, Neil Bezodis Orcid Logo, Huw Summers Orcid Logo, Paul Rees Orcid Logo

  • 59326.pdf

    PDF | Version of Record

    © 2022 White et al. This is an open access article distributed under the terms of the Creative Commons Attribution License

    Download (2.13MB)

Abstract

External peak power in the countermovement jump is frequently used to monitor athlete training. The gold standard method uses force platforms, but they are unsuitable for field-based testing. However, alternatives based on jump flight time or Newtonian methods applied to inertial sensor data have no...

Full description

Published in: PLOS ONE
ISSN: 1932-6203
Published: Public Library of Science (PLoS) 2022
Online Access: Check full text

URI: https://cronfa.swan.ac.uk/Record/cronfa59326
Tags: Add Tag
No Tags, Be the first to tag this record!
first_indexed 2022-02-08T08:49:11Z
last_indexed 2022-05-28T03:34:12Z
id cronfa59326
recordtype SURis
fullrecord <?xml version="1.0" encoding="utf-8"?><rfc1807 xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xmlns:xsd="http://www.w3.org/2001/XMLSchema"><bib-version>v2</bib-version><id>59326</id><entry>2022-02-08</entry><title>Determining jumping performance from a single body-worn accelerometer using machine learning</title><swanseaauthors><author><sid>725158b503e2be11ce4cc531afe08990</sid><firstname>Mark</firstname><surname>White</surname><name>Mark White</name><active>true</active><ethesisStudent>false</ethesisStudent></author><author><sid>534588568c1936e94e1ed8527b8c991b</sid><ORCID>0000-0003-2229-3310</ORCID><firstname>Neil</firstname><surname>Bezodis</surname><name>Neil Bezodis</name><active>true</active><ethesisStudent>false</ethesisStudent></author><author><sid>a61c15e220837ebfa52648c143769427</sid><ORCID>0000-0002-0898-5612</ORCID><firstname>Huw</firstname><surname>Summers</surname><name>Huw Summers</name><active>true</active><ethesisStudent>false</ethesisStudent></author><author><sid>537a2fe031a796a3bde99679ee8c24f5</sid><ORCID>0000-0002-7715-6914</ORCID><firstname>Paul</firstname><surname>Rees</surname><name>Paul Rees</name><active>true</active><ethesisStudent>false</ethesisStudent></author></swanseaauthors><date>2022-02-08</date><deptcode>SMA</deptcode><abstract>External peak power in the countermovement jump is frequently used to monitor athlete training. The gold standard method uses force platforms, but they are unsuitable for field-based testing. However, alternatives based on jump flight time or Newtonian methods applied to inertial sensor data have not been sufficiently accurate for athlete monitoring. Instead, we developed a machine learning model based on characteristic features (functional principal components) extracted from a single body-worn accelerometer. Data were collected from 69 male and female athletes at recreational, club or national levels, who performed 696 jumps in total. We considered vertical countermovement jumps (with and without arm swing),sensor anatomical locations, machine learning models and whether to use resultant or triaxial signals. Using a novel surrogate model optimisation procedure, we obtained the lowest errors with a support vector machine when using the resultant signal from a lower back sensor in jumps without arm swing. This model had a peak power RMSE of 2.3 W·kg-1 (5.1% of the mean), estimated using nested cross validation and supported by an independent holdout test (2.0 W·kg-1). This error is lower than in previous studies, although it is not yet sufficiently accurate for a field-based method. Our results demonstrate that functional data representations work well in machine learning by reducing model complexity in applications where signals are aligned in time. Our optimisation procedure also was shown to be robust can be used in wider applications with low-cost, noisy objective functions.</abstract><type>Journal Article</type><journal>PLOS ONE</journal><volume>17</volume><journalNumber>2</journalNumber><paginationStart>e0263846</paginationStart><paginationEnd/><publisher>Public Library of Science (PLoS)</publisher><placeOfPublication/><isbnPrint/><isbnElectronic/><issnPrint/><issnElectronic>1932-6203</issnElectronic><keywords/><publishedDay>10</publishedDay><publishedMonth>2</publishedMonth><publishedYear>2022</publishedYear><publishedDate>2022-02-10</publishedDate><doi>10.1371/journal.pone.0263846</doi><url/><notes/><college>COLLEGE NANME</college><department>Mathematics</department><CollegeCode>COLLEGE CODE</CollegeCode><DepartmentCode>SMA</DepartmentCode><institution>Swansea University</institution><apcterm>SU Library paid the OA fee (TA Institutional Deal)</apcterm><funders>The authors received no specific funding for this work.</funders><lastEdited>2022-07-13T12:23:18.1699680</lastEdited><Created>2022-02-08T08:43:43.0579993</Created><path><level id="1">College of Engineering</level><level id="2">Sports Science</level></path><authors><author><firstname>Mark</firstname><surname>White</surname><order>1</order></author><author><firstname>Neil</firstname><surname>Bezodis</surname><orcid>0000-0003-2229-3310</orcid><order>2</order></author><author><firstname>Jonathon</firstname><surname>Neville</surname><order>3</order></author><author><firstname>Huw</firstname><surname>Summers</surname><orcid>0000-0002-0898-5612</orcid><order>4</order></author><author><firstname>Paul</firstname><surname>Rees</surname><orcid>0000-0002-7715-6914</orcid><order>5</order></author></authors><documents><document><filename>59326__22355__00ce67b96c0e4d948043e611686b1b3f.pdf</filename><originalFilename>59326.pdf</originalFilename><uploaded>2022-02-11T09:22:38.3305038</uploaded><type>Output</type><contentLength>2234745</contentLength><contentType>application/pdf</contentType><version>Version of Record</version><cronfaStatus>true</cronfaStatus><documentNotes>© 2022 White et al. This is an open access article distributed under the terms of the Creative Commons Attribution License</documentNotes><copyrightCorrect>true</copyrightCorrect><language>eng</language><licence>http://creativecommons.org/licenses/by/4.0/</licence></document></documents><OutputDurs/></rfc1807>
spelling v2 59326 2022-02-08 Determining jumping performance from a single body-worn accelerometer using machine learning 725158b503e2be11ce4cc531afe08990 Mark White Mark White true false 534588568c1936e94e1ed8527b8c991b 0000-0003-2229-3310 Neil Bezodis Neil Bezodis true false a61c15e220837ebfa52648c143769427 0000-0002-0898-5612 Huw Summers Huw Summers true false 537a2fe031a796a3bde99679ee8c24f5 0000-0002-7715-6914 Paul Rees Paul Rees true false 2022-02-08 SMA External peak power in the countermovement jump is frequently used to monitor athlete training. The gold standard method uses force platforms, but they are unsuitable for field-based testing. However, alternatives based on jump flight time or Newtonian methods applied to inertial sensor data have not been sufficiently accurate for athlete monitoring. Instead, we developed a machine learning model based on characteristic features (functional principal components) extracted from a single body-worn accelerometer. Data were collected from 69 male and female athletes at recreational, club or national levels, who performed 696 jumps in total. We considered vertical countermovement jumps (with and without arm swing),sensor anatomical locations, machine learning models and whether to use resultant or triaxial signals. Using a novel surrogate model optimisation procedure, we obtained the lowest errors with a support vector machine when using the resultant signal from a lower back sensor in jumps without arm swing. This model had a peak power RMSE of 2.3 W·kg-1 (5.1% of the mean), estimated using nested cross validation and supported by an independent holdout test (2.0 W·kg-1). This error is lower than in previous studies, although it is not yet sufficiently accurate for a field-based method. Our results demonstrate that functional data representations work well in machine learning by reducing model complexity in applications where signals are aligned in time. Our optimisation procedure also was shown to be robust can be used in wider applications with low-cost, noisy objective functions. Journal Article PLOS ONE 17 2 e0263846 Public Library of Science (PLoS) 1932-6203 10 2 2022 2022-02-10 10.1371/journal.pone.0263846 COLLEGE NANME Mathematics COLLEGE CODE SMA Swansea University SU Library paid the OA fee (TA Institutional Deal) The authors received no specific funding for this work. 2022-07-13T12:23:18.1699680 2022-02-08T08:43:43.0579993 College of Engineering Sports Science Mark White 1 Neil Bezodis 0000-0003-2229-3310 2 Jonathon Neville 3 Huw Summers 0000-0002-0898-5612 4 Paul Rees 0000-0002-7715-6914 5 59326__22355__00ce67b96c0e4d948043e611686b1b3f.pdf 59326.pdf 2022-02-11T09:22:38.3305038 Output 2234745 application/pdf Version of Record true © 2022 White et al. This is an open access article distributed under the terms of the Creative Commons Attribution License true eng http://creativecommons.org/licenses/by/4.0/
title Determining jumping performance from a single body-worn accelerometer using machine learning
spellingShingle Determining jumping performance from a single body-worn accelerometer using machine learning
Mark White
Neil Bezodis
Huw Summers
Paul Rees
title_short Determining jumping performance from a single body-worn accelerometer using machine learning
title_full Determining jumping performance from a single body-worn accelerometer using machine learning
title_fullStr Determining jumping performance from a single body-worn accelerometer using machine learning
title_full_unstemmed Determining jumping performance from a single body-worn accelerometer using machine learning
title_sort Determining jumping performance from a single body-worn accelerometer using machine learning
author_id_str_mv 725158b503e2be11ce4cc531afe08990
534588568c1936e94e1ed8527b8c991b
a61c15e220837ebfa52648c143769427
537a2fe031a796a3bde99679ee8c24f5
author_id_fullname_str_mv 725158b503e2be11ce4cc531afe08990_***_Mark White
534588568c1936e94e1ed8527b8c991b_***_Neil Bezodis
a61c15e220837ebfa52648c143769427_***_Huw Summers
537a2fe031a796a3bde99679ee8c24f5_***_Paul Rees
author Mark White
Neil Bezodis
Huw Summers
Paul Rees
author2 Mark White
Neil Bezodis
Jonathon Neville
Huw Summers
Paul Rees
format Journal article
container_title PLOS ONE
container_volume 17
container_issue 2
container_start_page e0263846
publishDate 2022
institution Swansea University
issn 1932-6203
doi_str_mv 10.1371/journal.pone.0263846
publisher Public Library of Science (PLoS)
college_str College of Engineering
hierarchytype
hierarchy_top_id collegeofengineering
hierarchy_top_title College of Engineering
hierarchy_parent_id collegeofengineering
hierarchy_parent_title College of Engineering
department_str Sports Science{{{_:::_}}}College of Engineering{{{_:::_}}}Sports Science
document_store_str 1
active_str 0
description External peak power in the countermovement jump is frequently used to monitor athlete training. The gold standard method uses force platforms, but they are unsuitable for field-based testing. However, alternatives based on jump flight time or Newtonian methods applied to inertial sensor data have not been sufficiently accurate for athlete monitoring. Instead, we developed a machine learning model based on characteristic features (functional principal components) extracted from a single body-worn accelerometer. Data were collected from 69 male and female athletes at recreational, club or national levels, who performed 696 jumps in total. We considered vertical countermovement jumps (with and without arm swing),sensor anatomical locations, machine learning models and whether to use resultant or triaxial signals. Using a novel surrogate model optimisation procedure, we obtained the lowest errors with a support vector machine when using the resultant signal from a lower back sensor in jumps without arm swing. This model had a peak power RMSE of 2.3 W·kg-1 (5.1% of the mean), estimated using nested cross validation and supported by an independent holdout test (2.0 W·kg-1). This error is lower than in previous studies, although it is not yet sufficiently accurate for a field-based method. Our results demonstrate that functional data representations work well in machine learning by reducing model complexity in applications where signals are aligned in time. Our optimisation procedure also was shown to be robust can be used in wider applications with low-cost, noisy objective functions.
published_date 2022-02-10T12:23:16Z
_version_ 1738236385654996992
score 10.887713