One of mankind’s most major senses is its eyesight. The eye is different from the other body parts that make up the human’s sensor array. It is different, as through the eyes, a lot can be read with regards to a human’s expressions. For example, it can be assumed that a person’s attention is generally focussed on where they are looking and because of this, tracking the eyes movement can be useful. Fiction has had people operating things using sight rather than hands a long time ago, yet the technology hasn’t developed into the mainstream as of yet. The major reason seems to be the cost of systems, with most applications of gaze tracking being for specialist fields.

The development of this project is an attempt at making optical sensor control more feasible. The system is flexible enough to provide a basis for a variety of sensor applications, ranging from statistics or human-computer interfaces to safety control systems. It is also much cheaper than similar off-the-shelf products.

A simple Logitech WebCam is used to capture images and Infrared Light Emitting Diodes are mounted around the lens. Infrared light, reflected off the pupil, has been used to create a high intensity bright spot. This is called a Purkinje image and can be tracked. Edge Detection is used in conjunction with error and blink checking to derive an approximate corresponding x, y co-ordinate within the frame boundaries. Due to the possible effects of Infrared on the eye, the light source has been designed to Australian Standard. For the purposes of demonstrating the system, a MATLAB user interface has been designed. The interface provides a sight controlled telephone interface for hands free telephone dialling. The system is also easily adaptable to other potential Gaze Tracking uses. It has been written in MATLAB and operates in semi real-time, except in screen calibration, where single image pictures will be used.

There are a variety of ways this thesis may be extended upon. A Direct Show implementation could be used in conjunction with speed improvements to allow smoother tracking. Also, the system is reasonably intrusive, as one eye has to be covered up by the camera. Head tracking technology could be used with Gaze Tracking to allow a camera to be placed further away from the head, negating this problem.

Thesis Document (PDF)

Movie Download

Other Related Files

Poster Presentation (PDF)