Skip to Main content Skip to Navigation

Contribution to a non intrusive long-term sleep monitoring based on wearable patches and an algorithmic classification method

Qiang Pan 1
1 LAAS-S4M - Équipe Instrumentation embarquée et systèmes de surveillance intelligents
LAAS - Laboratoire d'analyse et d'architecture des systèmes
Abstract : Considerable effort has been devoted to academic and industrial research and development on wireless body networks for sleep monitoring in terms of non-intrusiveness, portability and autonomy. After an extensive review of current research on innovative technological systems and algorithms for sleep monitoring, this work presents three main contributions in sleep monitoring: - The implementation of a complete hardware architecture based on an IoT network. Movements and temperature data are collected from wearable autonomous devices (chest, wrist, feet), and sound, luminosity and room temperature data are collected from ambient sensors. This data is automatically sent to a remote database for display on a dashboard. - The proposal of two original methods for sleep stages classification (threshold-based methods and k-means clustering). In this work, the proposed algorithms use only the non-dominant wrist acceleration data. The computations lead to a classification into 4-sleep stages ("awake", "light sleep", "deep sleep" and "REM") for night sleep. The methods were validated with reference to the results obtained by two commercial devices "Fitbit" and "Withings Sleep Analyzer" and the subjective comments of the volunteers on their feelings about their sleep quality. Changes in sleep quality were evaluated between different nights with two volunteers to verify the performance of the algorithms. - The proposal and definition of sleep indicators to describe the sleep state and its quality via the calculation of a sleep score based on the duration of each sleep stage. Five volunteers were recruited for the tests during 15 nights and the performances between the two proposed algorithms were compared against the results of the "Fitbit" device. In terms of sleep stage classification, the device was compared to the clinical gold standard (PSG Polysomnography) on one subject during one night at the sleep clinic of the Purpan hospital in Toulouse. This work showed that it was possible to propose a light, non-intrusive, autonomous system for continuous sleep monitoring at home.
Complete list of metadata
Contributor : Laas Hal-Laas <>
Submitted on : Wednesday, April 7, 2021 - 3:37:18 PM
Last modification on : Friday, April 9, 2021 - 3:12:39 AM


PAN Qiang.pdf
Files produced by the author(s)


  • HAL Id : tel-03191905, version 1


Qiang Pan. Contribution to a non intrusive long-term sleep monitoring based on wearable patches and an algorithmic classification method. Micro and nanotechnologies/Microelectronics. Institut National des Sciences Appliquées de Toulouse, 2021. English. ⟨tel-03191905⟩



Record views


Files downloads