Physiological signal-based drowsiness detection using machine learning: Singular and hybrid signal approaches

Md Mahmudul Hasan; Christopher N. Watling; Grégoire S. Larue

doi:10.1016/j.jsr.2021.12.001

Back

Physiological signal-based drowsiness detection using machine learning: Singular and hybrid signal approaches

Journal article

Peer reviewed

Physiological signal-based drowsiness detection using machine learning: Singular and hybrid signal approaches

Md Mahmudul Hasan, Christopher N. Watling and Grégoire S. Larue

Journal of Safety Research, Vol.80, pp.215-225

2022

DOI: https://doi.org/10.1016/j.jsr.2021.12.001

PMID: 35249601

Files and links (1)

url

https://doi.org/10.1016/j.jsr.2021.12.001View

Published Version

Abstract

Accuracy

Drowsiness

Features

Ground truth

Machine learning

Physiological signals

Sensitivity

Specificity

Introduction: Drowsiness is one of the main contributors to road-related crashes and fatalities worldwide. To address this pressing global issue, researchers are continuing to develop driver drowsiness detection systems that use a variety of measures. However, most research on drowsiness detection uses approaches based on a singular metric and, as a result, fail to attain satisfactory reliability and validity to be implemented in vehicles. Method: This study examines the utility of drowsiness detection based on singular and a hybrid approach. This approach considered a range of metrics from three physiological signals – electroencephalography (EEG), electrooculography (EOG), and electrocardiography (ECG) – and used subjective sleepiness indices (assessed via the Karolinska Sleepiness Scale) as ground truth. The methodology consisted of signal recording with a psychomotor vigilance test (PVT), pre-processing, extracting, and determining the important features from the physiological signals for drowsiness detection. Finally, four supervised machine learning models were developed based on the subjective sleepiness responses using the extracted physiological features to detect drowsiness levels. Results: The results illustrate that the singular physiological measures show a specific performance metric pattern, with higher sensitivity and lower specificity or vice versa. In contrast, the hybrid biosignal-based models provide a better performance profile, reducing the disparity between the two metrics. Conclusions: The outcome of the study indicates that the selected features provided higher performance in the hybrid approaches than the singular approaches, which could be useful for future research implications. Practical Applications: Use of a hybrid approach seems warranted to improve in-vehicle driver drowsiness detection system. Practical applications will need to consider factors such as intrusiveness, ergonomics, cost-effectiveness, and user-friendliness of any driver drowsiness detection system.

Details

Title: Physiological signal-based drowsiness detection using machine learning: Singular and hybrid signal approaches
Authors: Md Mahmudul Hasan (Corresponding Author) - Queensland University of Technology
Christopher N. Watling (Author) - Queensland University of Technology
Grégoire S. Larue (Author) - Queensland University of Technology
Publication details: Journal of Safety Research, Vol.80, pp.215-225
Publisher: Elsevier Ltd
DOI: 10.1016/j.jsr.2021.12.001
ISSN: 1879-1247
PMID: 35249601
Organisation Unit: University of the Sunshine Coast, Queensland; School of Law and Society; Road Safety Research Collaboration
Language: English
Record Identifier: 99679186502621
Output Type: Journal article

Metrics

47 Record Views

21 Times Cited - Web of Science

InCites Highlights

These are selected metrics from InCites Benchmarking & Analytics tool, related to this output

Web Of Science research areas: Ergonomics; Public, Environmental & Occupational Health; Social Sciences, Interdisciplinary; Transportation

UN Sustainable Development Goals (SDGs)

This output has contributed to the advancement of the following goals:

Source: InCites