Monocular Gaze Tracking And Eye Behavior Identify

As an important organ for people to obtain information from the environment, eye is the main channel for the information to get into people's brain. The application of eye movement measurement is very extensive,which can be used in commercial, military, entertainment,daily life, human engineering and other fields. In recent years, in the field of computer vision and pattern recognition, the research on eye movement is becoming more and more popular. It will liberate people's hands when using gaze tracking technology into human-computer interaction system.Many methods of past research about eye tracking were designed to process high resolution images of eyes, so it cannot be well applied in many common scenarios. Research on monocular camera for gaze tracking is worth further study.The main results of this paper are as follows:1) In this paper, we pre-processed the images captured from monocular camera, including image filtering and image enhancement.Then we used Adaboost algorithm to detect the face and eyes. In order to accurately locate the eye center, for the people who wears glasses, we proposed a method based on active shape model to extract the eye glasses.Also, we added a step to roughly locate the eyeball for the people who does not wear glasses. It will improve the accuracy of eye center location in next step.2) In this paper, we use improved Hough circle detection, least squares ellipse fitting and improved location method based on curvature information to realize the eye center location and accurately obtain the position of the eye center.3) the obtained eye center coordinates are the input information of the gaze tracking. We use kNN and SVM algorithms to classify and recognize the sight trajectory. Also we detect the closed eyes and blink. The results of recognition and detection can be used as output signal in many practical scenarios.The glasses extraction algorithm can accurately obtain the glasses,and this algorithm is also suitable for other practical applications about the extraction of glasses. The proposed eye center location and trajectory recognition method based on monocular camera can accurately get the eye center and recognize different eye trajectories and behavior. The system is an improvement for the existing tracking system. The results of track recognition can be applied in a variety of applications.