next up previous contents
Next: 2.1 Techniques Based on Reflected Light Up: Eye-Gaze Media Previous: 1 Introduction

2 Present-day Eye-Gaze Tracking Techniques

Today, several ways of tracking the direction of eye-gaze exist. None of these techniques are perfect in the sense that no single technique fully satisfies all the usability requirements that Scott & Findlay (1993) cite Hallett (1986) for:

The ideal tracking device must

  1. Offer an unobstructed field of view with good access to the face and head
  2. Make no contact with the subject
  3. Meet the practical challenge of being capable of artificially stabilising the retinal image if necessary
  4. Possess an accuracy of at least one percent or a few minutes of arc; e.g. not give a 10° reading when truly 9°. Accuracy is limited by the cumulative effects of nonlinearity, distortion, noise, lag and other sources of error
  5. Offer a resolution of 1 minute of arc · sec¯ ¹, and thus be capable of detecting the smallest changes in eye position; resolution is limited only by instrumental noise
  6. Offer a wide dynamic range of one minute to 45° (= 3000-fold) for eye position and one minute arc · sec¯ ¹ to 800 · sec¯ ¹ (= 50,000-fold) for eye velocity
  7. Offer good temporal dynamics and speed of response (e.g. good gain and small phase shift to 100Hz, or a good step response).
  8. Possess a real-time response (to allow physiological manoeuvers).
  9. Measure all three degrees of angular rotation and be insensitive to ocular translation
  10. Be easily extended to binocular recording
  11. Be compatible with head and body recordings
  12. Be easy to use on a variety of subjects

While most people would agree that all these requirements are desirable, we must note that they are not all prerequisites for acceptable eye-gaze tracking interfaces. E.g., requirement b might not be crucial, as long as the physical construction of the device is discreet; as an example of today, one often finds the personnel of large companies wearing ID-cards for interfacing with electric doors etc. In the future, people might find it natural to wear, say, special glasses or `walkman'-ear-plugs, given that they are discreet and comfortable. One must also bear in mind that techniques that rely on head-mounted devices-be they inelegant headband-mounted cameras or handy glasses-have at least two advantages over most techniques that track at a distance: they track (almost) exactly the light rays that actually enter the eye and one tracker can cover an entire room. Thus, one can potentially make a camera-recording of exactly what the subject has looked at, instead of relying on several distant trackers and geometrical computations of directions of gaze for calculating the object that is being focused. This distinction becomes important if there is a possibility for some opaque object (e.g. a piece of paper) to obscure the screen, but not the tracking device (see figure 1). The dual-Purkinje-image tracking technique can in fact record the user's accomodation of focus, i.e., how far away the user is looking; thus it could probably reduce the "interfering problem," but this technique also requires the user's head to stay quite still in relation to the measuring equipment, and one way of doing this is by ... head-mounting!

Figure 1: An opaque object obscuring the display but not the distantly placed eye-gaze tracker might cause the tracker to compute the wrong point of regard.

While requirement j would be nice, as it would enable a greater freedom of movement for the user, allow one eye to be closed and generally yield more reliable tracking data, it is not essential for the average usage.

If we are to classify the current techniques of today by the way they make contact with the subject, there are basically three types of tracking techniques:

  1. Measuring the reflection of some light that is shone onto the eye. Typically, infrared light is used to distract the user as little as possible, and to avoid interference from other light sources like lamps.
  2. Measuring the electric potential of the skin around the eyes.
  3. Applying a special contact lens that facilitates tracking of its position.
Of these, technique type 1 includes techniques that make the least contact, and type 3 includes techniques that typically make the most contact.

All techniques require some form of calibration before usage, as well as some recalibration every now and then during use.


2.1 Techniques Based on Reflected Light
2.2 A Technique Based on Electric Skin Potential
2.3 Techniques Based on Contact Lenses
2.4 A Comparison of Tracking Techniques


next up previous contents
Next: 2.1 Techniques Based on Reflected Light Up: Eye-Gaze Media Previous: 1 Introduction

Authors: Arne John Glenstrup and Theo Engell-Nielsen