Sleep Monitoring Data Validation using a Linear Discriminant Classifier

Jose Perez-Macias Martin, Michael Pavel, Ilkka Korhonen, Holly Jimison

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


    Sleep quality has an important impact in
    overall health. There are several new commercially
    available devices for sleep tracking. Although they have
    been tested in sleep laboratories, for some users,
    estimates are reliable whereas for others are very
    different. In this paper we present a framework to
    predict the error range of the total sleep time. We
    studied 23 subjects’ sleep patterns using Fitbit One and
    Beddit Pro devices. We used Beddit Pro as our
    reference devices to verify Fitbit’s reliability. Previous
    analyses found that for some users the agreement
    between systems was high, whereas for others was low.
    We used a linear discriminant to predict this behavior
    in in the Fitbit data. 72.7% accuracy was achieved. Our
    results suggest that reliability of sleep estimates could
    be estimated without the use of a second device.
    Original languageEnglish
    Title of host publication7th annual Canadian Student Conference on Biomedical Computing and Engineering and IEEE EMBS International Student Conference (ISC)
    Number of pages2
    Publication statusPublished - 25 Jun 2014
    Publication typeA4 Article in conference proceedings
    EventIEEE EMBS International Student Conference (ISC'14 Oshawa) - University of Ontario Institute of Technology (, Oshawa, Canada
    Duration: 24 Jun 201426 Jun 2014


    ConferenceIEEE EMBS International Student Conference (ISC'14 Oshawa)
    Abbreviated titleCSCBCE / ISC
    Internet address


    Dive into the research topics of 'Sleep Monitoring Data Validation using a Linear Discriminant Classifier'. Together they form a unique fingerprint.

    Cite this