The Study Of Human-computer Interaction Operating Interface Based On Eye-gaze Tracking

Human-computer interaction(HCI)is to study the information exchange between human and computer.HCI is a comprehensive subject related but not limited to ergonomics,cognitive psychology,virtual reality and multimedia.We demonstrate the interface of HCI based on eye tracking methods using desktop real-time tracking binocular stereo vision(BSV).To meet the demand of sensitivity of this project,object detection technology,basing on dynamic background,provides data in real-time for the operator to select,determining the objects via eye-gaze tracking(ET),for the tracking operation.The desktop binocular ET model based on spatial state of pupil is proposed.Pupil center pointing method based on low gray value distribution is adopted with benefits of high point accuracy,fast calculation and non-threshold.The pupil edge detection method based on the radial derivative polar is proposed to improve the detection robustness and accuracy.Moreover,the pupil ellipse fitting is carried out by Random Sample Consensus(RANSAC),obtaining the pupil spatial information via the BSV and the parameters,to get the coordinates of the line of sight.An HCI model coordinating the sight-line is proposed,using the ORB to extract and describe the character point.Combining the Fast Library for Approximate Nearest Neighbors(FLANN)and the RANSAC to frame matching subtraction,the detected target images are obtained by filtering the connected fields.In the detection process,the closed-loop threshold adjustment is introduced in,which makes the model applicable to different environment,improving the system robustness.The experiment results manifest that the maximal errors of ET is 2.6° by proposed method.This system can provide 12 objects simultaneously,and the frame rate is 25 fps.To meet the requirement of real-time detection with high accuracy and robustness,this method provides an effective solution for the sensitive aiming system based on ET.