Vision Modelling. Part Four. Eye tracking

1 Excursion into the eye — 2 Perception — 3 Geometry of vision — 4 Eye tracking — 5 How to catch the gaze — 6 Modelling eye tracking

A post about the nature of what is actually being modelled, why this eye tracking is needed at all, and so on. If some of the terms are unclear, look at the third post on the geometry of vision.

Attention and saccades

Saccades can be controlled — they can be evoked consciously (by shifting the eyes) — but reflex movements are also possible, caused by the need to examine in more detail an object seen by peripheral vision. This is easy to notice: if something makes a sharp movement within peripheral sight, a reflex saccade will occur with the aim of examining more closely what caused that movement.

It is not hard to notice that the duration of the eyes’ optical focusing after a saccade on some region of an image is proportional to the attention devoted to a detailed study of that region.

Even before the appearance of eye-tracking technology, other, more laborious ways of tracking the traces were proposed (a little mirror on a contact lens and a spot of light thrown onto a photographic plate, as well as the method of magnetic-resonance tracking proposed by Academician Yarbus). Here is an example of a tracing from Yarbus’s monograph The Role of Eye Movement in the Process of Vision:

The duration of each recording is 3 minutes. Each fragment of the line corresponds to a saccade.

1. Examination of the painting without preliminary instructions
2. The task was given to assess the material circumstances of the family in the painting
3. Determine the age of the people depicted
4. Imagine what the members of the family were doing before the arrival of the unexpected guest
5. Remember how the characters are dressed
6. Remember the arrangement of people and objects in the room
7. Determine how long the unexpected guest had been absent

It is precisely because of the possibility of determining the degree of attention to particular regions of images that eye-tracking technology enjoys such great attention among designers and usability psychologists.

There is also a feedback between optical focus and attention: everything that finds itself in the eye’s field of view automatically begins to enjoy greater attention. Thus, if as a result of a reflex, unconscious saccade the optical focus has settled on some region (something “just pulls the eye”!), that region will be subjected to better study and will be better remembered.

Of certain interest as well is the question of where the end point of a saccade will lie, in the case of a conscious movement of the eyes toward a region of interest.

Unconscious movements

Unconscious saccades occur in accordance with four rules, each of which may play its part:

1. Accompanying motion.
   • A saccade is performed in order to anticipate the movement of the object being tracked. An example is the eye movements of people looking out of the window of a train.
   • Another possible cause that provokes this type of saccade is that the image of a potential movement has been caught. Example: an image of an unstable construction has been received. The gaze will reflexively jump to the calculated point where, in theory, that construction ought to fall.
   • Because, in a flat image, the gaze often tries to find and catch motion, a saccade of this type may also be provoked by the direction of the gaze in a person’s portrait, by the image of a flying arrow, and so on.
2. A psychologically active image has been seen by peripheral vision. For example, when peripheral vision detects any hint of a human face, a saccade will be performed in order to check that image. Besides people’s faces (for the recognition of which an entire dedicated part of the brain is responsible!), a psychologically active image might become, for a soldier, say, the image of a Kalashnikov rifle, and so on. ;)
3. A visually active image has been detected by peripheral vision. For example — a large object against a background of small ones, or emptiness against a uniformly painted field, or a bright colour patch. The saccade is performed for a more detailed examination.
4. A structurally more complex object has been detected by peripheral vision (composed of a larger number of clearly distinguishable contours, and so on) than the one being studied at the moment. The saccade is performed for a more detailed examination.

Unconscious movements of the first two types are difficult to model (except, perhaps, for the possibility of recognising a face or the presence of familiar symbols of text), since their necessity depends on the psychic patterns of image processing.

On the other hand, the latter two types of saccades are perfectly computable in the process of computer modelling. The last point is all the more important because it is precisely the last two rules that exert a considerable influence on the end point of both conscious and unconscious saccades.

The pupil and attention

It is interesting that there is also a positive correlation between the attention devoted to the study of some region, the duration of the gaze’s fixation on that region, and the width of the pupil. Experiments have shown that when intellectual effort increases (for example, the reading of some ridiculously complicated word, or a sharp emotional reaction) the pupil has a tendency to dilate.

In the case of modelling eye tracking, the program has no way of tracking the dilation of the pupil caused by the psychological components of the reaction to a picture, but it can estimate the duration and the degree of attention to particular regions that are difficult to perceive.

1 Excursion into the eye — 2 Perception — 3 Geometry of vision — 4 Eye tracking — 5 How to catch the gaze — 6 Modelling eye tracking

Published: 20 April 2009, 00:20
Author: ymik