Autoassociative Gaze Tracking System based on Artificial Intelligence

Abstract

Real-time auto-associative interface between the users gaze and a computer is presented and analyzed in this paper. This interface can provide more self-sufficiency to the disabled person and may help them while dealing with the problem of public integration. That is the reason for creation of real time and free positioning gaze tracking system which is applicable to control the computer application. The gaze tracking precision, computer processing rate and robustness of the system were explored experimentally. The artificial neuron network method and principal components analysis are used in the presented system for the user gaze and computer screen auto-association. The applied methods reduce the amount of the received video data by filtering out unimportant information either reduce the total computation burden of the system. Proper structure of neuron network and the number of the principal components were estimated through heuristic approach. The presented system of gaze tracking was tested with computer applications in real-time.