Brightness- and Motion-Based Blink Detection for Head-Mounted Eye Trackers

Blinks are an indicator for fatigue or drowsiness and can assist in the diagnose of mental disorders, such as schizophrenia. Additionally, a blink that obstructs the pupil impairs the performance of other eye-tracking algorithms, such as pupil detection, and often results in noise to the gaze estimation. In this paper, we present a blink detection algorithm that is tailored towards head-mounted eye trackers and is robust to calibration-based variations like translation or rotation of the eye. The proposed approach reached 96,35% accuracy for a realistic and challenging data set and in real-time even on low-end devices, rendering the proposed method suited for pervasive eye tracking.