Fatigue Detection Method Based on Smartphone Text Entry Performance Metrics

Al-Libawy, Hilal, Al-Ataby, Ali, Al-Nuaimy, Waleed ORCID: 0000-0001-8927-2368, Al-Taee, Majid A ORCID: 0000-0002-3252-3637 and Al-Jubouri, Qussay (2016) Fatigue Detection Method Based on Smartphone Text Entry Performance Metrics. In: 2016 9th International Conference on Developments in eSystems Engineering (DeSE), 2016-8-31 - 2016-9-2.

Text Fatig_Det.pdf - Author Accepted Manuscript Download (1MB)

Abstract

Workplace fatigue increases the risk of injuries or other accidents, and there is growing interest in developing means for the early detection of the signs of exhaustion. The ubiquity of smartphone use with integrated sensors allows the possibility of detecting the early signs of fatigue. Typing, chatting, Internet surfing and track-screen gestures are the most common tasks done by smartphone users in a daily manner. These tasks can be used to detect human fatigue in a non-intrusive way. This paper presents a human fatigue detection method based on virtual keyboard timing metrics which is calculated from smartphone text entry. This is achieved by using smartphone application that records keystroke events time and uses them as metrics to identify whether the user is fatigued or alert. Text entry error rate is considered as a type of psychomotor measures for fatigue/ alertness is utilized as a ground-truth metric in this study. A binary classifier based on support vector machine (SVM) classifier is suggested to identify the fatigue/alertness status of users participated in this study. The obtained results showed that the suggested method is highly accurate. The promising findings will facilitate development of a low-cost and non-intrusive mobile instrument for fatigue/alertness detection.

Item Type:	Conference or Workshop Item (Unspecified)
Uncontrolled Keywords:	fatigue, behavioral biometrics, smartphone, machine learning, error rate, SVM
Depositing User:	Symplectic Admin
Date Deposited:	19 May 2017 14:51
Last Modified:	19 Jan 2023 07:04
DOI:	10.1109/DeSE.2016.9
Related URLs:	Author Publisher
URI:	https://livrepository.liverpool.ac.uk/id/eprint/3007562