Eye Detection And Gaze Tracking

The eyes are important biological information, including the expression information, identity information, gender information, etc. It takes an important role in pattern recognition, fatigue driving detection and human-computer interaction. The paper mainly discusses eye detection and gaze tracking algorithm.On the basis of Paul Viola's face detection conception with rectangular features, this paper finds some rectangular features that is suitable with eyes, and with these features the Adaboost cascade classifiers are trained for eye detection. Then with the characteristics of symmetry of the eyes some of the geometric characteristics are adopted for correction. The geometric characteristics improve the accuracy of the eyes detection, and make the rough cascade classifier trained by few samples become a reality in application. After eye detection the eye blink detection and gaze tracking can be further studied. This paper briefly discussed eye blink detection and focusing on the issue of gaze tracking.The gaze tracking algorithm of this paper combines the space coordinates of the eye with the LBP characteristics of the infrared eye as the gaze direction features, used the support vector regression algorithm to predict the mapping function between the gaze direction and screen coordinates. The gaze algorithm can implement the human-computer interactive under natural head movement. The paper introduces a new binocular vision algorithm to calculate the space coordinates of the eye, this algorithm is simple, small amount of computation, more accurately, to meet the space coordinate calculation command in gaze tracking. Traditional infrared gaze tracking method usually takes pupil-glint vector. Although many people try various methods to accurately calculate the vector, pupil fuzzy borders, shape changes and other factors make the pupil center migration, and in addition, the reflection point sometimes appears to be too large that offsets the actual position, which caused the pupil-glint vector inaccurate and affect experimental results. Using LBP features of infrared eye images, not only the texture changes when gaze direction change can represent, but also includes the pupil-glint vector information. This method can avoid the disadvantage of the calculation of pupil-glint vector, and can estimate the gaze direction more accurately.