A Fast and Adaptive Video-Based Method for Eye Blink Rate Estimation

Eye blink rate (EBR) estimation is one of the informative cues and challengeable areas in eye-based systems that has a wide range of applications like the detecting a drivers drowsiness, anxiety analysis, diseases detection and etc. This paper presents an adaptive blink rate estimation algorithm. The advantages of this algorithm are simplicity, accuracy, fastness, low computational cost and robustness against lighting conditions. This algorithm is based on simple image processing techniques. The first step of blink detection method is eye detection. To accomplish this task, we suppose that a fairly large face image is available. Each frame of the input video is processed and the location of the eye is found. The next step calculates a value to determine the state of eye. Our method uses this value to EBR estimation. This paper presets the accuracy of new algorithm by providing a data set of several people and comparing the results with some of the strong relevant methods. The experimental results show that the proposed method has overall accuracy of 98.91%. The average blink rate estimation time of new algorithm for a sample is less than 80 micro seconds, which makes it suitable for real-time applications.