Pilot Study on Gaze-Based Mental Fatigue Detection During Interactive Image Exploitation

Human mental fatigue occurs during various tasks due to increased load and time-on-task. As it might impair human performance, it could be beneficial detecting it automatically and subsequently implement measures to mitigate fatigue. To accomplish this, mental fatigue has to be detected, preferably as unobtrusive as possible. Recent research proposes that remote eye-tracking could be a promising method. The background of this contribution is interactive image exploitation as it might occur in safety or security applications. We consider wide area motion imagery which typically covers several square kilometers and includes a huge number of tiny vehicles and persons. A human operator has to perform lots of search, zoom and pan operations in order to find relevant objects. We conducted a pilot study (N = 20 non-expert image analysts) where subjects preformed several basic image exploitation tasks. During the sessions, we collected their gaze data using a 500 Hz eye-tracker. From the recorded gaze data protocols, we extracted saccadic and fixational gaze parameters using the I-VT algorithm. Mean and maximum saccadic velocity as well as mean saccadic amplitude decrease over time. This corresponds to findings by the research community in terms of observed gaze behavior under mental fatigue. However, the effects are small and need confirmation by future work.