Study And Implementation Of Gaze Tracking System

Gaze tracking is a technique to determine the direction of gaze of human eyes by extracting eye features. Gaze tracking is a very important technology in human computer interaction, and can be applied in many fields, such as helping the disabled, advertising analysis, clinical medicine. In recent years, gaze tracking technology has achieved great development, but there are still some problems in concrete application, the shortcomings are as follows:(1) Some gaze tracking system have complex structure and with harsh conditions of using, which brings great inconvenience to users;(2) Some gaze tracking system can not reach the requirement of the actual use in the precision;(3) Some gaze tracking system response is slow, cannot achieve the real-time effect;(4) The majority of gaze tracking system needs to use various kinds of auxiliary equipment, such as multiple cameras, infrared light source, so the system hardware is expensive.In order to solve the above problem, this paper designed a single camera and no infrared light source eye gaze tracking system. The system adopts the pupil localization algorithm based on ellipse fitting, this algorithm can detect the iris center accurately in ordinary office illumination, and has good stability and frame rate, and enhanced the performance of the system, reduced the demand for hardware; The system using artificial neural network method to map the eye image and the fixation point, and the structure of artificial neural network and feature vectors were improved and optimized, so that the system does not require the use of infrared light source and a plurality of cameras, reduced the system hardware requirements and the hardware cost, improved the tracking precision of the system and enhanced the practicability of the system.In this paper, the main work are:(1) The pupil localization algorithm were studied, through the pupil localization algorithm based on ellipse fitting, detected the iris center of the eye accurately, and puts forward a method based on gradient vector to locate the pupil center, this method can detect the eye pupil center accurately;(2) The eye gaze tracking algorithm were studied in this paper, the gaze tracking methods based on the artificial neural network are improved and optimized, improve the tracking precision;(3) Completed the system design and realization of hardware platform and system software.