Eye Tracking Technology Research Based On Desktop

We human beings get information mainly through our eyes, it has a great significance to study how we make use of the human eye visual information for research. The human eye tracking research is considered to be the most effective way in the study of visual information processing. Currently, eye tracking technology has been widely used in medicine, psychology, human-computer interaction and many other fields.According to the differences of hardware structure, the current mainstream of eye tracking system can be divided into two kinds:head-mounted eye tracking system and remote eye tracking system. This paper mainly analyzes the technology principle and method used in remote eye tracking system. The primary research contents includes:face and eye detection, the localization of iris center, and the calibration of the human eye fixation point. The main works and research results are summarized as follows:1. Face detection and localizationThis paper uses skin model and Adaboost algorithm for face detection. Firstly, the image is converted into YCbCr color space, secondly, the values of Cb, Cr is used to get the face candidate regions, and finally the Adaboost algorithm is used in the face candidate regions to detection face.2. Eye detection and localizationThis paper first summarizes the eye detection algorithm in common use, the advantages and disadvantages are given. Then we improved the existing algorithm, we found that the grey level of human eye region is lower than other parts of human face region, using this character, we can obtain the candidate region of human eye, and then take advantages of following information:the size of human eyes and the position of human eyes in face, we can remove the interference candidate regions and locate the true eye position.3. Localization of iris centerIn order to locate the iris center, the iris boundary is detected first. This paper uses histogram equalization, image binarization, morphology processing and edge detection to get iris boundary. After get the boundary of iris, the least square method is used to ellipse fitting the boundary, the center coordinate of the elliptic curve is the center of the iris.4. Calibration of eye gazeThe calibration of eye gaze is a very important step in eye tracking system, it gives the real position when people look at the computer screen. We assume that the coordinates of iris center in the eye image and the coordinates of fixation point on the screen has a one-to-one mapping relationship, and this paper uses the mapping relationship to calculate fixation point. Before using the eye tracking system, let the user look at calibrated points on the screen, the computer can records the corresponding coordinates of iris center, using this data to get the mapping relation between iris center coordinate and fixation point coordinate. Then we can use this mapping relation to get the coordinate of fixation point. During the process of calibration, the collected data may has noise, if we use the data with noise to estimate mapping function parameter, it may cause the fixation point calibration error, in this paper, we introduce a random sample consensus (RANSC) method to solve the problem.