Deep link provided by Citebite
Close this shade
Source:  http://www.diku.dk/~panic/eyegaze/node22.html

4.3 Summary of Reported Uses of Eye Tracking

Eye-gaze tracking has been used in laboratory experiments for several decades, but as yet the techniques have not been developed much for non-laboratory tasks. However, a few actual applications for human-computer interaction with real-time analysis of eye-gaze tracking data have been constructed.

4.3.1 "The Little Prince" Storyteller

The idea of this application is taken from Antoine de Saint Exupery's novel "The Little Prince" which also has been turned into an animated film for children, where the evolution of the world and the boy's (mental state) is expressed. The story takes place in outer space on a planet inhabited by a little prince. The little planet (it is approximately 200 feet across) has several features: volcanoes, staircases (as a hobby the little prince collects staircases) and flowers that orbit the planet.

"The Little Prince" storyteller system utilizes the corneal/pupil reflection principle using an infrared camera placed near an infrared light source and samples every 16 milliseconds (60 Hz). Starker & Bolt (1990) write that the system makes distinctions between focusing on a single object and focusing on groups of the same type of objects, i.e. when looking at a certain staircase you will receive information on this staircase, and when looking at a group of staircases, you receive general information on staircases. The system has three types of interest level algorithms:

1. When gazing at an object, a counter for this object is incremented. The counter is the computed level of interest. This gives problems if the system must evaluate whether the user is losing interest in the subject and should return to a higher level.
2. The interest level is calculated with regard to the elapsed time using the formula:
   interest level = c1 × counter - c2 × elapsed time (c1 and c2 are predefined constants.)
3. This algorithm makes the level of interest in all objects smaller by multiplying it with a calculated constant. Then a "fresh look constant" is added to the object looked at. The next subject will then be 1) A specific object, if the interest level is concentrated on this one object. 2) Similar objects, treated as a category, if more than one object of the same type have high interest level. 3) The world in general, when no specific object or group of objects has been found. In this way a system has been made that, instead of using a simple average to find the subject of interest, tries to find out on which level the user is interested in the object(s) gazed at.

4.3.2 Applications for Disabled People

One of the most frequent uses of eye-gaze techniques-apart from laboratory experiments-has been in assistive devices for severely disable people. Chapman (1991) reports having used an eye-gaze system for letting people suffering from the so-called locked-in syndrome communicate with their surroundings and control their environment (electrical appliances). The system used is based on the corneal/pupil reflection relationship technique, which is well-suited for disabled, as it is non-intrusive and because the requirement to keep the user's head still (unfortunately) is fulfilled in this case.

The system is calibrated in the standard way by letting the user successively look at nine points on the screen. The system consists of a main menu from which the user can select one of several controlling screens to typewrite, place a telephone call, control a TV-set, call upon the caregiver and turn appliances on or off. All this is naturally done by looking at the screen; all selections are effectuated after the user has gazed at the selection for a latency period. Chapman (1991) reports that 75% of locked-in syndrome patient can significantly benefit from using the eye-gaze system.

Frey et al. (1990) have implemented a similar system for eye-gaze-response interface computer aid, Erica. The main difference is that due to limited resolution of the eye-gaze, the display of an entire keyboard is replaced with six large eye-gaze operated keys. This, combined with access to common phrases and employment of a character-prediction algoritm using Markov chains for changing the virtual menu structure dynamically, decreases eye-typewriting time by 25%

4.3.3 Eye-Controlled Camera Focusing

Recently Canon has developed eye tracking techniques for their cameras. First the Canon EOS 5 (Can 1992) was introduced, then the Canon UC-X1 video camera (HIFI elektronik 1995).

Canon EOS 5

The canon EOS 5 camera is one of the first cameras to offer eye-gaze control using a CCD censor in the camera's ocular. The gaze direction is found while the user is looking the ocular and the nearest of 5 "eye-select" points shown through the ocular is selected and the camera adjusts its features (e.g. focus) accordingly. The camera has memory for 5 individual calibations and Canon claims that everybody can calibrate and use the camera. We believe this cannot be factual, since we encountered at least one person at the Experimentarium who could not use the EyeCatcher due to the geometry of her eyeball.

Canon UC-X1 Hi

The world's first eye-gaze controlled video camera is also manufactured by Canon. The autofocus system is based on the system used in the Canon EOS 5 camera (explained above). Whereas the Canon EOS 5 camera is limited to 5 focus points, the UC-X1 Hi video camera has no such limitations. It simply focusses on the object gazed at. The camera uses four infrared light sources, of which two are used for people with glasses and are placed further from the eye than the other two (somehow the video camera is capable of detecting whether the user wears glasses or not). The calibration is done by looking at two dots on the video display of the ocular. In the review in HIFI elektronik (1995) they have found several problems when using the camera: During normal use the video camera can all of a sudden fail to track the eye properly, which forces the user to make sudden movements with the eyes to regain correct tracking. A white square is constantly displaying where the user is looking. We think this must be a very annoying feature, since you cannot avoid looking at this "fly" hovering in front of your eyes all the time.