What does the eye tracker measure?

Nowadays, the rapid development of IT and the emergence of many interesting gadgets, many have heard of such a thing as an eye-tracker (this is a device that determines where people are looking) and even saw various kinds of "heat maps" and analytics built "on the basis of data eye-tracker. But at the same time, very few people think about what exactly this device measures ...

Very clearly it was seen during the discussion, which unfolded around my previous note with a bunch of illustrations about the difference in the perception of sites. All very heatedly discussed the advantages and disadvantages of different types of design; how can I interpret the maps I have cited; Is it enough to use 5 respondents or 10? whether such tests are needed at all ... But no one wondered what data was displayed on the “attention maps” I quoted!

So let's first determine what data eye-tracker gives out, and then talk about their usefulness in real life.
')

Let's start with the device itself. In contrast to the electrodes and helmets that appear to many, the modern eye-tracker looks almost like a regular monitor, in which several cameras and infrared lights are mounted.

During operation, the camera continuously takes your face, highlights it on the frames of the eye, and by triangulation determines the position of each eye in space relative to the eye-trackera. In addition, by the relative position of the center of the pupil and glare from the IR illumination, the direction of gaze is determined (for each eye). Well, since the eyes of different people are somewhat different, then in order for this direction to be determined accurately, a calibration is carried out before starting the test: the eye-tracker shows you a moving ball, you follow it with a look, and the gauge calculates all sorts of factors that you need to use when working with you. And, finally, knowing the position of the eye relative to the screen and the direction of the gaze, that point of the screen that you are currently looking at is calculated.


Thus, several dozen times per second, the eye-tracker records data about the subject — the location of each eye, the coordinates of the eyes of each eye, the quality of measurements, the resulting coordinate of the eyes (ideally, they should coincide and the closer the points of view of each eye are, the more reliable the measurement ), timestamp and some other parameters. At the same time, the course of the test is recorded - what exactly is shown on the screen and what actions the subject performed: he pressed the keys, moved the mouse, etc. Both of these data streams are synchronized in time.

Simultaneously with the recording of this whole affair, a calculation of the views takes place. As far back as the 19th century, it became known that a person’s gaze does not move continuously, but in spurts, moving from one point to another. The rapid movements of the gaze are called saccades, and the points where the gaze stops are fixed by fixing points or fixation. In Russian, most often under the "look" (in the context of eye-tracking'a) refers precisely to fixation.

During the recording, the eye-tracker keeps track of the coordinates of your gaze (I remind you that this happens several dozen times per second) and, if they remain unchanged, summarizes the time. When a certain threshold value is exceeded - usually 100 milliseconds - the device notes a fixation. The coordinates of the gaze, however, are also not completely unchanged: they can “tremble”, for example, because of rounding errors (we recalculate the pixels), because of the movements of the subject or because of blinking. Therefore, a certain threshold radius is taken (usually 30 or 50 pixels) and, if the coordinates remain inside a given circle, then the fixation continues, and if they exit, the countdown of the new fixation begins.

In the course of testing, the operator has the opportunity to observe his progress - on the second monitor, the materials displayed by the subject and the image of the subject's gaze superimposed on them are displayed in real time.

As a matter of fact, it is the sequence of fixations that is the result of the work of the eye-tracker. Then begin the mapping, analysis and interpretation of this sequence. But we will discuss this in detail in the next post.