Fast Detection and Classification of Drivers’ Responses to Stressful Events and Cognitive Workload

Abstract

We apply machine learning techniques to detect moments of stress and cognitive load during simulator driving experiences. The use of the electrical skin conductance, or more precisely the electrodermal activity (EDA), is particularly interesting for assessing drivers’ states because it is easily measurable; it is also involuntary and uncontrollable. Detection of responses to external stimuli can be performed on a scale of seconds with an accuracy of 86%. Moreover, we observe that responses to stress events and cognitive efforts can be differentiated with an accuracy of 80% over sub-minute time intervals. We compare our results to others reported in the literature. Automatic and fast detection of responses to stressful events and high cognitive workload can be used to assess drivers’ user experience (UX) and their interaction with their vehicle.