Ever since remote eye trackers started to appear on the market, manufacturers have looked for ways to measure and report eye tracker technical specifications that could be used to describe their performance. However, this process was done independently by each manufacturer and no standards have yet been established. Consequently, the technical specifications for eye trackers are often difficult to compare, with each manufacturer providing a value that describes a specific attribute without clearly defining it, nor stating the methodology used to measure it.
The goal of this talk is to examine closely the accuracy and precision specifications of an eye tracker manufacturer (Tobii Technology AB). The rational of the test method behind these specifications is to provide a standardized method to measure different remote eye tracker models, in different user relevant scenarios. The method tests a variety of factors that affect the eye tracker performance, such as the test room lighting conditions, head position in the track box and stimulus angle placement on the screen. Precision is also measured on artificial eyes in order to differentiate between system noise and expected precision on human eyes, with human artifacts included. Accuracy and precision are calculated from stable eye tracking data (high track robustness individuals). This latest test requirement has an important effect on the way the eye tracker specifications should be interpreted and extrapolated to other eye tracking setups.
I will finish this talk by briefly discussing future improvements and changes to this method as well as the inclusion of other important eye tracking metrics.