Detection of Elevated States of Cognitive Demand in Drivers in a Naturalistic Driving Environment: Feasibility Study using Support Vector Machines and CTOF-Framework

DOI: 10.5176/2251-2179_ATAI29

Authors: Zach Chuanzhong Tan, Bryan Reimer, Bruce Mehler and Joseph F. Coughlin

Abstract:

The ability to detect states of elevated cognitive demand with relatively high accuracy is a fundamental requirement for the development of useful workload management systems. Using a secondary cognitive task to establish a period of known high demand in a sample of 74 drivers, attributes drawn from physiology, eye behavior and driving performance were examined individually and in combination using machine learning algorithms to assess relative sensitivity in detecting elevated demand under actual highway driving conditions. While there are good reasons to expect that there are conditions under which eye behavior and driving performance metrics provide important information, for the pure cognitive demand condition studied, physiological measures showed marked superiority in detecting elevated cognitive demand.

Keywords: cognitive demand; mental workload, distraction, support vector machines; machine learning; driver state estimation; physiology; eye tracking; workload manager; driving safety