Detecting Human Eye Blinks through OpenCV
Kirti Hemant Wanjale, Araddhana Arvind Deshmukh, Jyoti Chetan Vanikar, Alpana Prashant Adsul, Shailesh Pramod Bendale · 2024
Blink detection, a crucial part of facial movement analysis, is applied in a number of fields, including human-computer interaction, driver monitoring, and healthcare. However, automatic blink recognition presents unique obstacles due to variability in blink rates and ambient conditions. This work proposes a novel real-time method to robustly and accurately detect eye blinks in video sequences using OpenCV and the Modified Eye Aspect Ratio (Modified EAR) as a crucial component. The suggested methodology's automatic facial landmark detectors are initially trained on a range of real-world datasets. When it comes to adapting to shifting surroundings, such as shifts in illumination, expressions on faces, and head orientations, these detectors are incredibly versatile. Our method carefully calculates the locations of facial landmarks with every frame in the video sequence, making it possible to calculate the vertical distance between the eyelids with accuracy. This metric, which takes the form of the Modified EAR, serves as a single scalar quantity that represents the degree of eye closure or openness. Our experimentation's empirical results highlight the facial landmarks' remarkable precision in reliably assessing the dynamics of eye opening and closure. Blinks are detected by applying a threshold value to the Modified EAR, which enables the detection of blink patterns that take place within a short temporal window. Our approach not only detects blinks efficiently but also outperforms current state-of-the-art methods, as demonstrated by the results obtained from a large and thorough dataset.