Detecting Mental Fatigue in Vessel Pilots Using Deep Learning and Physiological Sensors

Abstract

Nowadays, human related issues are the main causes of accidents in the maritime domain. Among these issues, mental fatigue is responsible for reducing cognitive capabilities, situational awareness, and decision-making skills. Early detection and assessment of mental fatigue can be used to reduce the number of causalities, to the benefit of crewmembers, ship owners, and the maritime environment. Although the use of physiological sensors is the most trusted approach for measuring mental fatigue, it is a complex task due to the different ways mental fatigue can manifest in different people. In this paper, we present the application of deep learning techniques and physiological sensors to assess mental fatigue in the maritime domain, using a vessel piloting task as case study. The results demonstrate that because of their ability to extract features otherwise hard to recognize from in data, deep learning techniques in special convolutional neural networks can achieve high levels of mental fatigue classification accuracy, although cross-subject classification performance is still not sufficient for real-life applications.