Eye tracker accuracy terms and definitions - working copy

1 reply [Last post]
mulvey
mulvey's picture
Offline
Joined: May 25 2010

Here you can view and comment on a draft set of definitions and terminology which the technical committee are working on right now. We will update in November according to the work of the committee.

Thanks for participating!

Fiona Mulvey
Lund University, Sweden
COGAIN Chair

mulvey
mulvey's picture
Offline
Joined: May 25 2010
Discussion points currently considered by the TC

Points for discussion (April 2010) - Kenneth Holmqvist

- Position of the eye in camera
- mathematical calculation of accuracy and precision vary: many test procedures may be necessary to look at different aspects of inaccuracy
-calibration
-- eye-shapes
-- instruction
-- precise criterion for accepting calibration
-- is the selected set of calibration targets to be the same for all tested eye-trackers?
---- many different types of accuracies vs. "off-the-shelf" accuracy with all the default settings- Software development may quickly change values of a specific eye-tracker
--- If any manufacturer thinks they have a better product now than when first tested, send it to an independent lab to verify it. The standard should also describe methods of in house testing which can aide the design process.
- accuracy across the monitor - how to deal with variant accuracy across the screen/display
- accuracy throughout the headbox
- where does the system start and end? if the data is smoothed by the system, should we consider the system a 'black box' and work with whatever data is reported? Defining 'the system' exactly
-- For comparative testing of eye trackers, is there some way of presenting exactly the same eye and eye movements to more than one eyetracker? what if we bypass the camera, and feed several system with the same video of an eye?
-- the system as a black box
- filters - ignore them? take them into account?
-- raw data vs. smoothed by the systems vs. smoothed by the researcher in a MatLab filter
- 100 people versus an artificial eye
- A multidimensional space, where dimensions are accuracy, precision, sampling frequency
-- Both eye-trackers and the requirements of application areas span limited areas inside the space
-- Sampling frequency is a highlighted property; but there is no corresponding competetion for the most precise or accurate
-- Most journal papers do not report measured accuracy and precision - only reiterate manufactures flyers

Fiona Mulvey
Lund University, Sweden
COGAIN Chair