Real-Time Eye Gaze Tracking With an Unmodified Commodity Webcam Employing a Neural Network

摘要

An eye-gaze-guided computer interface could enable computer use by the seriously disabled but existing systems cost tens of thousands of dollars or have cumbersome setups. This paper presents a methodology for real-time eye gaze tracking using a standard webcam without the need for hardware modification or special placement. An artificial neural network was employed to estimate the location of the user's gaze based on an image of the user's eye, mimicking the way that humans determine where another person is looking. Accuracy measurements and usability experiments were performed using a laptop computer with a webcam built into the screen. The results show this approach to be promising for the development of usable eye tracking systems using standard webcams, particularly those built into many laptop computers.