Region Of Interest Tracking On The Screen Based On Gaze Detection

With the development of computer science and the requirement of human computer interaction application, the application of human-computer interaction has attracted more and more researchers’ attention. Visual system not only plays an important role for people to obtain information, but also reflects the great significant information of human responding to the outside world. Therefore, in the study of human-computer interaction, eye tracking is an important research. Commercial gaze tracking instrument can be applied in some professional fields with high accuracy; however, it can’t be widely used in circumstance such as the user attention analysis, advertising analysis and education application because it is expensive. Nevertheless, analyzing user behavior based on a webcam can be achieved without special equipment. The aim of this study is to design a gaze tracking system based on a common camera under natural light, and the main achievements are introduced as follows:1、In this paper, gradient information is used to locate the iris center accurately, by optimizing the coefficient matrix and the objective function. Compared with many existing algorithms, our algorithm has better effect to locate the iris center, providing reliable data for the following research.2、In order to build a real-time system and optimize the performance, Parameterized Appearance Models based on Supervised Descent Method is used to get facial feature points, and precisely locate the eye area. Since head movement has a great influence on gaze detection, which in practice is inevitable, the head rotation angles are also detected and considered as input date of the gaze mapping function.3、In this paper, the feature data including accurate position of the pupil, the eye contour points and head motion parameters is used as the input data and locate the user’s interest area by machine learning method. SVM (Support Vector Machine) is applied respectively in horizontal and vertical direction, and different kernel function and parameters are adopted to improve the precision.This paper proposed a stable and reliable algorithm to locate users’area of interest based on single camera when they look at the screen. It is designed to obtain the regions of attention without head-mounted device, multiple cameras or infrared technology. Compared with the existing gaze tracking system, our system is a kind of supplement, with the characteristics of low cost, convenient, concise, and easy to popularize.