Eye Tracking System For Human-Computer Interaction On Mobile Devices

Eye tracking technique could analyze and record the fixation point coordinates,which have been widely used in psychology,advertising evaluation,human-computer interaction and other fields.With the rapid development of computer vision,image processing technology and the improvement of hardware,eye tracking has been applied to mobile devices.However,due to the limited hardware performance of mobile devices,such as low resolution of camera,eye tracking for mobile devices still have low computation efficiency but high error rate,compared to personal computers.In order to implement the eye tracking system for mobile devices,an improved eye tracking approach was proposed,and the related human-computer interaction application was also biuld.The main research work is desecibed as follows:(1)Image processing approaches for mobile device's eye tracking.Due to lack of hardware resources of mobile devices,the eye tracking efficiency was low.Therefore,the image processed was devied into multiple steps,from general to detailed,and the image processing efficiency was improved.Firstly,an improved local binary feature cascade classifier was used to detect the face of the images,and then the face region was extracted.Then,the Haar feature cascade classifier was used to detect the image of human eyes,and the eye region was extracted finally.(2)Calculating fixation point for mobile device's eye tracking.Because the human eye region detection bias and other errors,after the human eye region was extracted,the human eye region was expanded to obtain more accurate pre-selected region(including the integrity of the eye contour,but compared to human pre selected region more accurately remove the hair and eyebrows disruptors).Then the binarization and contour extraction was used to obtain accurate eye circumscribed rectangle,and iris and upper eyelid connectivity region contour.The eye center(EC)and the center of gravity(CG)of iris and upper eyelid connectivity region contour was calculated.The connected vector(EC-CG vector)from the center of eye image circumscribed rectangle to the gravity center of iris and upper eyelid contour was established.At last,the mapping relation between EC-CG vector and the fixation point was confitured by the calibration method,and calculated the accurate fixation point.(3)Human-computer interaction application research based on mobile device's eye tracking.The eye tracking prototype system based on smart phone was designed and developed.The system could obtain the eye tracking data accurately and efficiently.On this basis,the eye tracking data visualization and analysis was proposed.The hot map,the marked image of important areas of interest and the gaze transition image between areas of intresest.Finally,used the advertisement images on smart phones as the example,and allowed multiple users sharing their eye tracking data visualization.The accuracy of eye tracking approach and the prototype application system have been tested.Eye tracking precision user test results show that the tracking accuracy of this paper is 2.80 to 3.56 degrees of angle,and the average processing speed of eye tracking is 6.01 frames per scond.It shows that eye tracking appraoch has high accuracy,and has a huge role in promoting eyetracking tracking for mobile devices.Protorype application system user test results showed that the efficiency of information retrieval and the level of comprehension were improved by the eye tracking data visualization.It validated the feasibility and effectiveness of the mobile device based eye tracking technique and human-computer interaction application in this paper.