Implementation Of Eye-Tracking Technology And Its Application In The Research Of Joint Attention

Joint attention refers to a behavior that someone's sight is synchronized with others' and pays attention to the same thing,to achieve the purpose of sharing an interest in that thing.It is found that the ability of joint attention has developed well when the children are about the age of three.Children can use this ability to learn a variety of knowledge and experience.Having this ability not only helps children adapt to the environment better,expand the ways and means of children's understanding of the world,and communicating with others but also helps children develop the ability of speech,emotion,self-awareness,and so on.However,not all children can develop a well-coordinated ability of joint attention around the age of three.Some children can be found that their ability of joint attention is defective through observation when they are about three years old.In order to improve this kind of children's ability of joint attention,it is necessary to carry out intervention training for these children,and the way of intervention training is mostly invasive.Because invasive intervention train-ing will cause children's uncomfortable feelings and make children appear psychological conflict,this thesis proposes to use eye-tracking technology to design a non-invasive joint attention training system.In this thesis,eye-tracking technology is studied in-depth,and four main steps of this technology are implemented:face detection,eye location,pupil center location,and gaze estimation.For face detection,this thesis proposes to use the Haar classifier to detect the face.Firstly,Haar-like feature template is used to calculate the feature values;Secondly,the AdaBoost algorithm is used to train the weak classifiers;Thirdly,a group of strong classifiers is trained according to these weak classifiers.Finally,the strong classifiers are cascaded to judge the face or non-face.For eye location,this thesis proposes to use the method of prior knowledge based on "three courts and five eyes" and the method of gray-level integral projection to implement the location of the eye area,which solves the problem of interference of eyebrow for pupil center location.For the pupil center location,this thesis proposes to use the method based on the gradient to implement the pupil center location.This algorithm mainly uses the large gap between the gray value of iris and the gray value of sclera,which results in a large gradient at the edge of the iris,and the extension line of all gradients will intersect at one point,which is the pupil center.For gaze estimation,this thesis proposes to use the mapping relationship between coordinates of the pupil center and coordinates of the fixation point to estimate the gaze.Firstly,using nine points to calibrate and a mapping function between the pupil center and fixation points can be obtained through the calibration process,and then by using this function can calculate the estimated fixation points,so as to estimate the gaze.After researching and implementing eye-tracking technology,this thesis proposes to ap-ply this technology to the research of joint attention.The primary method of implementation is to use a multi-threaded way to let a game of training joint attention and the eye-tracking program run simultaneously.There are two main processes shown to the subjects.One is a calibration process,which is implemented using eye-tracking technology,and the other is a small game to train the ability of joint attention.There is a virtual character in this game,and the training of the ability of joint attention can be achieved through the induction of the virtual character's gaze and head.After the experiment,visual eye movement data and digitized eye movement data will be obtained.By analyzing these eye movement data,it can judge whether the ability of joint attention of the subjects has been improved.