Development of an automatic drowsiness monitoring system using the electrocardiogram
Bourghelle, Florent
Promotor(s) : Verly, Jacques
Date of defense : 27-Jun-2016/28-Jun-2016 • Permalink : http://hdl.handle.net/2268.2/1451
Details
Title : | Development of an automatic drowsiness monitoring system using the electrocardiogram |
Translated title : | [fr] Développement d'un système de surveillance automatique de la somnolence à l'aide de l'électrocardiogramme |
Author : | Bourghelle, Florent |
Date of defense : | 27-Jun-2016/28-Jun-2016 |
Advisor(s) : | Verly, Jacques |
Committee's member(s) : | Geurts, Pierre
Phillips, Christophe Embrechts, Jean-Jacques Kolh, Philippe |
Language : | English |
Number of pages : | 81 |
Keywords : | [en] Drowsiness, ECG |
Discipline(s) : | Engineering, computing & technology > Electrical & electronics engineering |
Institution(s) : | Université de Liège, Liège, Belgique |
Degree: | Master en ingénieur civil électricien, à finalité approfondie |
Faculty: | Master thesis of the Faculté des Sciences appliquées |
Abstract
[en] The aim of this thesis consists of the development of an automatic drowsiness monitoring system based on the electrocardiogram (ECG). Moreover, as the feasibility of this physiological signal to detect drowsiness is still not proved, this thesis also investigates its feasibility.
This thesis is based on an experiment were subjects were sleep deprived during 28 hours. At 3 specific moments of sleep deprivation, subjects performed psychomotor vigilance task (PVT). During these tasks, different physiological signals whose electroencephalogram (EEG), electrooculogram (EOG), and electrocardiogram (ECG) were recorded. Based on the EEG and EOG signals, which are the references to assess drowsiness, the true state of each subject is known on the Karolinska Drowsiness Scale and can be defined as awake or drowsy given a defined threshold.
First, this thesis performs a review of the literature to find the possible parameters indicative of drowsiness computed from the ECG. Then, a complete processing chain of the ECG signal is implemented to be able to compute these parameters in the time and statistical domains, the non-linear domain, and finally in the frequency domain from the raw ECG of the subjects. As the respiratory signal can be derived from the ECG (ECG-Derived Respiration signal), this thesis also incorporates parameters from the respiratory domain in order to see if this domain can be use to detect drowsiness.
Once these parameters are computed, a machine learning phase is developed. During this phase, the issue of the variability of the features between the subjects was highlighted. Several techniques to compensate this variability have been tested but none improved the results obtained. This variability makes the system developed to be not reliable enough on all the subjects of the experiment to only use the ECG to predict drowsiness.
File(s)
Document(s)
Cite this master thesis
The University of Liège does not guarantee the scientific quality of these students' works or the accuracy of all the information they contain.