Research And Application Of Gaze Direction Calibration Technology

With the development of computer science and technology,human-computer interaction has made great progress.The line-of-sight tracking technology has been widely used in the field of human-computer interaction,due to its response speed block and convenient operation.The eye tracking method based on pupil-corneal reflex is favored by researchers.On the basis of domestic and foreign research,aiming at the problem of low accuracy of gaze estimation during head movement,a new method of eye direction calibration based on head motion and line-of-sight tracking is proposed.It can effectively improve the accuracy of gaze estimation when the head moves freely.This paper mainly completes the following work:(1)Face detection.For the problem that the detection result of a single face detection algorithm is not ideal,in this paper,Adaboost algorithm combined with ASM algorithm is used to detect and track faces.The number of face detections is reduced,the detection efficiency is improved,and efficient and accurate face detection is realized.(2)For the problem that the accuracy of the gaze point estimation is inferior,this paper proposes a multi-point calibration dynamic parameter calibration method to calibrate the gaze point estimation results.First,divide the screen into four rectangular areas,each of them does not interfere with each other,and calculate the optimal calibration parameters for each area.The calibration parameters of the fixation point are approximated by combining the optimal parameters with other marker point parameters.Then dynamically select the calibration parameters of the region where the fixation point is located to calculate the virtual projection point,and establish a projection plane.Finally,the coordinates of the fixation point are calculated by the invariance of the cross ratio.The method solves the problem that the gaze point estimation result shift caused by the pupil center and the facula spot are not in the same plane.(3)Aiming at the problem that the gaze point detected during the head movement is different from the actual gaze point,the cross-sectional invariant fixation point is estimated.In this paper,the head space attitude compensation method is used to reduce the influence of head motion on line of sight tracking.First,the POSIT algorithm is used to obtain the head pose information.Then,the head movement information and the eye movement information are integrated and the position of the fixation point is determined according to the actual situation of the movement of the head movement.Finally,the actual fixation point is converted to screen coordinates.The method establishes a position-independent line-of-sight tracking model,and the user can move at will without affecting the accuracy of the line-of-sight tracking,and realizes the line-of-sight tracking of the coordinated movement of the head and eye.This article conducted an accuracy test on eight testers.Experiments were carried out to verify the gaze point estimation results in the situation of relative static of the head and free movement of the head.When the head is relatively stationary,the traditional method of gaze point estimation is compared with the gaze point estimation result of the dynamic parameter calibration method proposed in this paper,and the average error is analyzed.When the head is relatively stationary,the traditional method of gaze point estimation is compared with the gaze point estimation result of the dynamic parameter calibration method proposed in this paper,analyze the respective average errors;When the head is free to move,the gaze point estimation result obtained by the traditional method is compared with the gaze point result obtained by the head compensation in this paper,analyze the respective average errors.The experimental results show that the estimation accuracy of the fixation point when the head is relatively stationary is increased from 2.8° to 1.1°,and the error range is reduced by 35.2%.The accuracy of gaze point estimation during head movement increased from 4.1° to 1.7°,and the error range was reduced by 58.2%.Through the analysis of experimental results,the line-of-sight calibration technique improves the calculation accuracy of the gaze point of free movement of the head,and confirms the effectiveness of the proposed method.