Subject recognition using wrist-worn triaxial accelerometer data

Stefano Mauceri, Louis Smith, James Sweeney, James McDermott

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

This study demonstrates how a subject can be identified by the means of accelerometer data generated through wrist-worn devices in the context of clinical trials where data integrity is of utmost importance. A custom vector of features extracted from the daily accelerometer time series is defined. Feature selection is adapted to take account of the sequential structure in features. Several classifiers are compared within three different learning frameworks: binary, multi-class and one-class. A simple algorithm like logistic regression shows excellent performance in the binary and multi-class frameworks.

Original languageEnglish
Title of host publicationMachine Learning, Optimization, and Big Data - Third International Conference, MOD 2017, Revised Selected Papers
EditorsGiuseppe Nicosia, Giovanni Giuffrida, Panos Pardalos, Renato Umeton
PublisherSpringer Verlag
Pages574-585
Number of pages12
ISBN (Print)9783319729251
DOIs
Publication statusPublished - 2018
Externally publishedYes
Event3rd International Conference on Machine Learning, Optimization, and Big Data, MOD 2017 - Volterra, Italy
Duration: 14 Sep 201717 Sep 2017

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume10710 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference3rd International Conference on Machine Learning, Optimization, and Big Data, MOD 2017
Country/TerritoryItaly
CityVolterra
Period14/09/1717/09/17

Keywords

  • Accelerometer data
  • Anomaly detection
  • Classification
  • Clinical trials

Fingerprint

Dive into the research topics of 'Subject recognition using wrist-worn triaxial accelerometer data'. Together they form a unique fingerprint.

Cite this