real-time eye blink detection using facial landmarks (paper)

The Objectives of a paper

SVM classifier detects eye blinks as a pattern of EAR values

real-time eye blink detection using facial landmarks.pdf

What is EAR(Summary)?

The eye aspect ratio (EAR) between height and width of the eye is computed.

reference : Real-Time Eye Blink Detection using Facial Landmarks

EAR function

EXPERIMENT

For observation, i insert code to see EAR history

arr.append(ear)
...
plt.plot(arr)
plt.show()

Graph of EAR Value Changes by my blink

we can know that if EAR is smaller than around 0.3~0.28, then blinking. --> EYE_AR_THRESH = 0.3

Additional

If you want more detailed study go to

https://www.pyimagesearch.com/2017/04/24/eye-blink-detection-opencv-python-dlib/

If you want source code

https://github.com/sunjungAn/Opencv_practice/blob/master/Facial%20landmarks%20with%20dlib%2C%20OpenCV%2C%20and%20Python/detect_blinks.py