Evaluating keystroke dynamics as a biomarker for mental fatigue detection

Abstract

Abstract In this paper we study the feasibility of employing keystroke biometrics for mental fatigue detection during natural typing. For this task, we employ TypeNet, a state-of-the-art deep neuronal network, originally intended for user authentication at large scale using keystroke dynamics. We adapted TypeNet for fatigue detection by leveraging the information embedded in TypeNet for person recognition, and applying that information to a different but related task as it is fatigue detection by employing domain adaptation techniques. All experiments were conducted using three keystroke databases that comprise different contexts and data collection protocols. Our preliminary results showed performances ranging between 72.2% and 80.0% for fatigue versus rested sample classification, which is aligned with previously published models on daily alertness and circadian cycles. This demonstrates the potential of our proposed system to characterize mental fatigue fluctuations via natural typing patterns. Finally, we studied the feasibility of an active detection approach that utilizes the continuous monitoring of keystroke biometric patterns for the real-time assessment of subject fatigue.