BLINK AND PUPIL BASED DETECTION AND CLASSIFICATION OF USER ACTIVITIES AND LIKENESS/DISLIKENESS TOWARDS THESE ACTIVITIES

Abstract

Science takes part to observe eyes behavior, and it and make things easier to do. The previous research shows that when a something the pupil dilate and when a human dislike something the pupil Generally, it is quite difficult to know someone’s feelings about likeness and ensure through eyes human like contracts. dis likeness towards specific activities. In this study pupil detections algorithm is divided into three main parts, capturing of eyes, segmentation of pupil and algorithm which fits circles around the iris and pupil to measure pupil/iris ratio. The objective of this project is to develop the system that detects which activity human on computer through tracking their blinking patterns, because blinking rates varies for different activities and likeness and dis likeness towards those activities by tracking user’s pupil size. This project is divided in two subsequent parts identifying blinking had first trace the face of is performing patterns and tracking pupil size. For blinking we rates or human using shape_predictor_68_face_landmarks.dat datasets and captures video from live video stream using cv2.VideoCapture() library’s object then we got the both left and right eyes landmarks, draw contours on the eyes and applied some more mathematical calculation to detect blinking of eyes. Finally, modules give 71.54% accuracy when was tested we had concluded by evaluating the results that properly, which is better than previous related researches