Drowsiness Detection A Blink-Based Approach Using Computer Vision Techniques

Real-time detection of drowsiness is an important need in efforts to improve safety, especially in the transportation sector. This research focuses on the development of a non-invasive system to detect drowsiness based on eye blink analysis. This system utilizes computer vision technology to measure blink frequency and duration as the main parameters. With an Eye Aspect Ratio (EAR)-based algorithm implemented, the system can detect changes in eye blink patterns accurately. The data obtained is used to identify signs of drowsiness such as increased blink duration or decreased frequency. The study evaluates hardware performance for real-time image processing in detecting drowsiness through blink dynamics. The results reveal that an AMD A9 laptop performs significantly better, achieving a sampling rate of 16 Hz, compared to the Raspberry Pi 3B+, which operates at 1.6 Hz, approximately 10 times slower. Additionally, the experiment establishes two drowsiness detection criteria: the duration of a single blink and the frequency of blinks within a set timeframe. The latter requires additional program functionality to store and analyze blink data over time. These findings underscore the need for robust hardware and software optimization to ensure reliable, real-time drowsiness detection. Experimental results show a high level of accuracy in detecting drowsiness, making it a reliable solution to improve user safety in various sectors.