A SURVEY ON REAL-TIME EYE BLINK DETECTION USING FACIAL LANDMARKS

In many contexts, such as facial movement analysis and signal processing, blink detection is a crucial approach. However, due of the blink rate, automatic blink detection is quite difficult. This study suggested a real-time technique for identifying eye blinks in a series of videos. Automatic facial landmark detectors show extraordinary resilience to a variety of environmental parameters, such as lighting conditions, face expressions, and head position, after being trained on a real-world dataset. The suggested method determines the locations of facial landmarks for each video frame and then uses those sites to extract the vertical distance between the eyelids. Our findings indicate that the identifiable landmarks are reliable enough to establish the consistency of eye opening and closure. The suggested approach calculates the positions of the facial landmarks, uses the modified eye aspect ratio (Modified EAR) to extract a single scalar quantity, and describes the closeness of the eyes in each frame. The Modified EAR threshold value and recognising eye blinks as a pattern of EAR values in a brief temporal window are the last methods for detecting blinks. The findings from a representative data set show that the proposed methodology is more effective than the most recent method.