EventIDE provides a built-in calibration procedure for all supported eye-tracker models. The built-in GLM calibration is recommended, even if you calibrate your eye-tracker with native manufacturer tools. The GLM calibration procedure uses the least square linear method to fit the eye-tracker gaze positions to pixel coordinates on the screen. The linear fitting is done separately for X and Y signals and results in estimation of two pairs of coefficients of a General Liner Model: X Gain, X offset and Y gain and Y offset. The coefficients are applied into GLM equations (shown below) that are later used in EventIDE during runtime eye-tracking.
The GLM calibration is a very accurate method, given that the raw tracker signal is closely linear. However, some eye-trackers models may operate with non-linear internal corrections of the data. Therefore, eye-tracker’s native calibration procedures are recommended to run, before repeating GLM calibration in EventIDE (e.g. Popup Calibration utility in EyeLink Display software package).
GLM Calibration Window
The calibration procedure in EventiDE has to be run for each participant before starting an experiment. Select the eye-tracker element in the EventIDE designer and browse its properties in the Property Grid. There is the GLM Calibration group that looks like this:
Notice that both, gains and offsets, are set to the default norm values, 1 and 0. Click on the play button in the Run GLM Calibration line. A new full-screen window will be open occupying the entire screen (in multi-monitor setups on the stimulus monitor selected in the EventIDE’s preferences) .
The calibration screen is automatically scaled to the screen resolution chosen for the current experiment (in Display Settings). The calibration runs, as you press the Start button on the right. On each step in a calibration process participant has to fixate on the calibration marker that moves in a new position. The calibration finishes automatically.
Validation of the calibration results
After the calibration finishes, you will see the results. The red points show the measured gaze positions at the location of each calibration point. The cumulative errors are shown in the bottom panel, both in pixels and visual degrees. There are 3 detailed validation modes that you can select on the panel.
As you start the calibration window, you see the calibration settings panel. If the settings are adjusted, new values is saved with the current experiment.
Improving calibration results
if you are getting a large mean pixel error in calibration, you can try several approaches to improve bad calibration results:
- Use a smaller Layout Scale – calibration locations will be closer to the screen center, in which eye-trackers perform better
- Increase the Fixation Time value in the settings. Then the calibration can be calculated better, due to large number of samples
Back to the designer
After closing the calibration window , the program updates the GLM coefficients on the tracker element. The Calibration group in the Property Grid should look like this:
Now, it’s time to run an experiment and enjoy reliable and accurate eye-tracking.