Feasibility of smart wearables for driver drowsiness detection and its potential among different age groups. Issue 1 (2nd January 2020)

• Record Type: Journal Article
• Authors: Kundinger, Thomas; Yalavarthi, Phani Krishna; Riener, Andreas; Wintersberger, Philipp; Schartmüller, Clemens
• Is Part Of: International journal of pervasive computing and communications. Volume 16:Issue 1(2020)
• Page Start: 1
• Page End: 23
• Publication Date: 2020-01-02
• Subjects: Wearable devices -- Physiological measures -- Driver drowsiness detection -- Advanced driver assistance systems (ADAS) -- Simulator study -- Active safety -- Driver monitoring -- Heart rate variability (HRV) -- Machine learning; Ubiquitous computing -- Periodicals; Mobile computing -- Periodicals; Computer network protocols -- Periodicals; Computer network architectures -- Periodicals; Application software -- Development -- Periodicals; 004.6
• DOI: 10.1108/IJPCC-03-2019-0017
• Languages: English
• ISSNs: 1742-7371

Abstract

Abstract : Purpose: Drowsiness is a common cause of severe road accidents. Therefore, numerous drowsiness detection methods were developed and explored in recent years, especially concepts using physiological measurements achieved promising results. Nevertheless, existing systems have some limitations that hinder their use in vehicles. To overcome these limitations, this paper aims to investigate the development of a low-cost, non-invasive drowsiness detection system, using physiological signals obtained from conventional wearable devices. Design/methodology/approach: Two simulator studies, the first study in a low-level driving simulator ( N = 10) to check feasibility and efficiency, and the second study in a high-fidelity driving simulator ( N = 30) including two age groups, were conducted. An algorithm was developed to extract features from the heart rate signals and a data set was created by labelling these features according to the identified driver state in the simulator study. Using this data set, binary classifiers were trained and tested using various machine learning algorithms. Findings: The trained classifiers reached a classification accuracy of 99.9%, which is similar to the results obtained by the studies which used intrusive electrodes to detect ECG. The results revealed that heart rate patterns are sensitive to the drivers' age, i.e. models trained with data from one age group are not efficient in detecting drowsiness for another age group, suggesting to develop universal driver models with data from different age groups combined with individual driver models. Originality/value: This work investigated the feasibility of driver drowsiness detection by solely using physiological data from wrist-worn wearable devices, such as smartwatches or fitness trackers that are readily available in the consumer market. It was found that such devices are reliable in drowsiness detection.