Method And Application Of Gaze Point Trajectory Description Based On Eye Movement Video

Eye tracking technology is a new type of Human-Computer Interaction(HCI)technology.And fixation point trajectory description method is the critical algorithm,whose accuracy directly affects the effect of eye tracking.By pupil location and fixation point calibrating on those eye movement videos,this method can figure out the mapping relationship between the pupil center and the fixation point,thereby estimating the fixation point of the subject,and obtaining the fixation point trajectory and the region of interest.The fixation point trajectory description technology has a wide application prospect in various fields including medical treatment,psychological cognition,personalized advertisement recommendation,and security.However,nowadays the domestic exploration of fixation point trajectory description technology is not popularized.Therefore,it is of great significance to accelerate the pace of research and improve core competence in this technology,thereby breaking the foreign countries' monopoly,promoting the modernization construction of national defense,and narrowing the gap between us and developed countries as soon as possible.This thesis presents a method for describing the fixation point trajectory based on eye movement videos,the major steps of which include collecting eye movement video data through a camera,preprocessing the video data frames,performing blink detection to filter out the blink frames,and then working out the pupil center coordinate,setting out the mapping relationship between the pupil center and the screen fixation point with the fixation point calibration,and finally calculating the fixation point in watching the screen according to the mapping relationship.The main content of this thesis is as follows:(1)This thesis analyzes the research status of the fixation point trajectory description system,and proposes an algorithm based on eye movement videos aimed at the existing problems.This method firstly uses an infrared camera to collect eye movement signals and performs some preprocessing like filtering and denoising.Then the blinking detection and pupil location are carried out,on the basis of which the pupil-fixation point second-order mapping relationship is established to realize the fixation point trajectory description.Owing to the data requirements of the algorithm,an experimental paradigm and three types of observed pictures in experiment were also designed to collect eye movement data in the video,and a small database of eye movement videos was initially established.(2)In the pupil location module,a method combining coarse pupil location and fine pupil location is presented.First,we use an infrared camera to collect eye movement signals and preprocess the images of eye movement through making binary image by adaptive method in order to improve the accuracy of the video data.Then,blinking detection is performed to filter out the blink frames,and we use the row and column location method to realize the coarse location of the pupils.We try to fit the pupil with an ellipse to find an approximately circular pupil area,and then use the threshold centroid method to locate the exact location of the pupil center.Experimental results show that the method is highly effective in finding the location of the pupil center point,and can largely improve the functioning performance of pupil location.(3)In the fixation point calibration module,this thesis uses a nonlinear second-order polynomial function to find the mapping relationship between the center location of human pupils and the coordinate of fixation point in the screen.Considering the difficulties in the calibration and the accuracy in solving the mapping relationship,this thesis applies the more efficient nine-point calibration method.The results indicate that in the laboratory environment,the average error of the fixation point description of eight subjects is 0.414 radian.Compared with the traditional fixation point description method(0.57 radian),the accuracy is increased by 0.156 radian.(4)The fixation point tracking description system based on eye movement videos is designed on the MATLAB platform.It is mainly composed of four functional modules: the data acquisition module,the pupil location module,the fixation point calibration module,and the result output module.People can manipulate the system through a simple graphical user interface.The system can efficiently support the optimization of advertising web layout,so as to help businesses to promote products to consumers in a targeted way,which has high potential application value.