Proposal for a Eye Blink Detection Using Mediapipe, Eye Aspect Ratio and Peak Identification

This study presents an algorithm to detect eye blinks in webcam video footage. The algorithm leverages the Mediapipe facial landmark detector to compute the Eye Aspect Ratio (EAR), quantifying the level of eye openness in each video frame. Blinks are identified using a peak detection technique. The algorithm’s efficacy was assessed using the Eyeblink 8 dataset, where it achieved an accuracy of 98.03% and a precision of 83.42%. Additionally, the algorithm extracts several blink-related parameters, including duration, number, and rate. These parameters were statistically analyzed and compared with findings from existing literature, showing no significant differences.