CONICET | Buscador de Institutos y Recursos Humanos

Automatic gaze estimation not based oncommercial and expensive eye tracking hardware so-lutions canenable several applications in the fields ofhuman-computer interaction (HCI) and human behav-ior analysis. It is therefore not surprising that sev-eral related techniques and methods have been investi-gated in recent years. However, very few camera-basedsystems proposed in the literature are both real-timeand robust. In this work, we propose a real-time user-calibration-free gaze estimation system that does notneed person-dependent calibration, can deal with illumi-nation changes and head pose variations, and can workwith a wide range of distances from the camera. Oursolution is based on a 3-D appearance-based methodthat processes the images from a built-in laptop camera.Real-time performance is obtained by combining headpose information with geometrical eye features to traina machine learning algorithm. Our method has beenvalidated on a data set of images of users in a naturalenvironment, showing promising results. The possibilityof a real time implementation, combined with the goodquality of gaze tracking, make this system suitable forvarious HCI applications.