Eye Tracker
	chil home > lab only > polocies and procedures >

CHIL Policies and Procedures: Eye Tracker

This section is written to give you some guidance in running an experiment on the CHIL eye tracker. It is divided into five sections: Apparatus/Overiew, Programming, Setting Up, Running an experiment, and Replay Scans.

1. Apparatus / Overview of Tracker
The eye tracker is an ISCAN RK726/RK520 HighRes Pupil/CR tracker with a Polhemus FASTRACK head tracker. Head-mounted optics and a sampling rate of 60 Hz are used. This system, like many other laboratory eye trackers, works by shining an infrared light on the eye and taking a video image of the eye. From that image, it is possible to determine the pupil center and the point on the cornea closest to the camera (the corneal reflection) and take the vector between them. This vector changes as the eye orients to different positions on the screen, and with calibration to known points, it is possible to compute visual point of regard (POR, also referred to as "point of gaze"). The magnetic polhemus is used to compensate for head movements. POR reports by the eye-tracking equipment are typically accurate to within one-half degree of visual angle. POR and mouse position are recorded approximately every 16.7 ms (corresponding to 60 Hz) by the experimental software. Stimulus and POR/mouse data for each trial are recorded so that all individual trials can be "replayed" at various speeds. An experimenter monitors each experimental trial and recalibrates the eye tracker if there appear to be a sizable disparity between reasonable expectations about where users were looking and the position reported by the tracker. (For a thorough discussion of the system and its components, there is a manual--a black 3-ring binder--that is worthwhile to look through before you begin. It's kept in the eye-tracking room.)

2. Programming an Experiment for the Tracker
Examples of experiments that were run on the tracker can be found on the hard drive there. Of particular interest is the file labeled, "EPAL-Menu". This is Mike's experiment, which tracked the eye movements of participants using pull-down menus (Byrne, Anderson, Douglass, Matessa, 1999). It is of particular interest because the code is well documented. Other examples of experiments are EPAL -Icon, EPAL-IconSpacing, and EPAL-Deception.
There are three primary files that will have to be altered to run an experiment on the Eye Tracker. All of them, stimulus-generation.lisp, stimulus-presentation.lisp, and global-interface-objects.lisp are in the "shared" folder of EPAL-Menu (likewise for each the other eye tracking experiments).
You'll want to create your own folder and naming system, such as EPAL-XXXX. (We'll use EPAL-XXXX as our example in the rest of this document.)

3. Setting up the Tracker
If you must start from scratch, actually attaching the cables to each of the relevant pieces of hardware, there are several diagrams in the black binder that will be useful. Notice that each of the cables and their hosts have been carefully labeled to make this process easier. For example, the cable labeled "Scene Camera Video Out" goes in the receptacle on the Scene Camera Control Unit also labeled "Scene Camera Video Out."
You will also need to configure the eye-tracking software. This involves defining the planar surface on which the eye movements will be tracked (the computer screen in this case, although multiple surfaces may be defined, such as the keyboard if necessary). All plane positions are specified with respect to the magnetic head tracker source, which must remain fixed within the environment.

Two other points: Position the Polhemus tripod such that the sensor on it is no more than 30" from where the participant's head will be. Also, it should be either directly behind the participant or to his/her right (the same side as the sensor on the head gear). I suggest marking three spots on the floor, one for each leg of the tripod, so that if the magnetic head tracker source is moved, you can easily put it back.

To define the planar surface coordinates, you will need to measure in inches the distance from the center of the polhemus at the top of the tripod to the four corners of the screen in x, y, and z coordinates. Note that the x-plane is vertical, the y-plane is horizontal (left/right), and the z-plane is the horizontal distance from the polhemus to the front of the screen. You'll be prompted to save these coordinates in a file - WRITE DOWN the name of this file. You'll need it later and it is hard to retrieve. (Detailed instructions on how to set up these configurations are contained in the black folder under the sections "SFT-LOS: Defining Planes in the Environment" and "SFT-LOS: Setting the System Configuration.")

4. Running the Tracker
The first step is to start up everything in the eye tracking room.

Here's a list of what that will be:
Dell PC labeled "ISCAN ETL-500 Eye Tracking Computer"
Macintosh PPC
3 Sony Monitors (VDU, Eye, and Scene)
Video Scan Converter Plus (Gray box)
ISCAN Eye Camera Control Unit (small black box)
ISCAN Scene Camera Control Unit (larger black box)Next, you will need to load the appropriate software.

Procedures for running an experiment
On the Mac:
-Start up the lisp environment (MCL) and load the experiment (load-system :epal-xxxx).
-If you've already run the experiment or part of it since restarting the system, type (dispose-epal-input) to reset.
-Under the menu "EPAL-XXXX" select globals/checkset
-Then select the appropriate files associated with your program. The first is a resource file for the images used in the experiment, the second is a file containing the processing regions associated with your experiment. You can select the POR regions file for your experiment if you have it at this point. You do not need this file to run the experiment, however, just to analyze the data. For Now, you can select "Actual" instead of your own POR regions file.
-Next select "EPAL-XXXX/Do it!/browse subjects." Here you can enter a new subject, continue with a previous subject, etc., according to the options on the pull-down menu.

On the PC:
-Run the file "ISCAN LOS1.50N"(lower icon of two icons in the upper right corner of the desktop)
-From the MAIN COMMAND MENU select
1. Define Planar Surface Coordinates for Environment
Select 1. Read Planar Surface Coordinates from disk file
Enter the name of the file you have created (see "Setting up the Tracker"), for example, "plane3.def". Follow the instructions on the screen.
-Return to the MAIN COMMAND MENU and select
2. Run HEADHUNTER System
The section describes the steps required in order to track a participant. A list of the key commands (available to you via the visible screen) and their functions is provided in the black binder under the section, "RK-620PC: System Power Up and Operation."
In order to track a participant, the system must be calibrated to the individual participant. Begin the calibration process by having the participant sit in a comfortable position just as they will sit during the experiment.
Now, place the headgear on the participant. Be careful with the glass pane on the headgear. Its edges are sharp, and cutting participants is bad. Important things to note when placing the headgear on the participant include:
-the headgear should be stable, i.e., doesn’t slip. If it slips, the subject will have to be recalibrated. It may help if the headgear rests on the crown of the participant’s forehead.
-The cable attached to the headgear should not hang freely. It is heavy and may cause the headgear to slip. To keep it from hanging, you might try clipping it to the back of the participant’s chair or having the participant sit on it. Whatever works.
Now you’ll want to adjust the optics (glass pane) so that you have a clear image of the entire eye on the monitor labeled “Eye Monitor.” This may require you to adjust the headgear as well.

Now hit the L key twice. This toggles the serial output of the system off and on.

Next, the pupil threshold level should be adjusted.
Hit F7 and use the up or down arrow keys to adjust. You want the "highlighted" dark region to cover as much of the pupil as possible—use the eye monitor—without highlighting dark regions outside of the pupil.

Next, the corneal reflection level should be adjusted.
Hit F8 and use the up or down arrow keys to adjust. You want the crosshairs representing the point of corneal reflection to be stable.
Have the participant look at each of the four corners of the screen. The crosshairs should remain on the center of the pupil and the point of corneal reflection as they move their eyes around the screen. If they do not, you may need to adjust the thresholds and/or the headgear.

Next, the subject must be "boresighted." Boresighting is the process of calibrating the head-tracking system. Essentially, the system must know where the participant’s head is (based on the sensor on the right side of the headgear) relative to the Polhemus (sensor on tripod), relative to the screen plane.

To boresight: Hit the R key to reset. Instruct the participant to hold their head still and to look at the center of the screen, then hit the G key to boresight. Immediately after hitting “G,” look at the monitor labeled “VDU Monitor” and note the values labeled “H” and “V”. These will pause briefly. When they pause, you want their values to be near to 255. Nearer is better (say plus or minus 5 or so, but you’ll have to use your own judgement). If they are not close, re-boresight. You can test the boresight process by looking at the VDU monitor. If everything has worked, there should now be a cursor in the crosshairs in the center of the screen. Have the participant move their head left to right. The cross hair should smoothly move in correspondence to their head movements. Same thing for nodding up and down. If it doesn’t move smoothly and correspondingly, you must re-boresight.

Finally, the subject’s eyes must be calibrated. The eye calibration involves having the participant look at locations on the screen that deviate by a known number of degrees of visual angle and having the system note the divergence in their pupil and corneal reflection points as they look at these known locations. The calibration box, in the lower-left corner of the PC monitor, contains a number which represents the number of degrees of visual angle that will be used in the calibration process. You may adjust this number with the up or down arrow keys. According to the folks at ISCAN, this number should be near to, or slightly less than, 5.

To begin the eye calibration: Hit R to reset. Then hit C for calibration. Instruct the participant that they must remain as still as possible for the next minute or so. If they move too much, they will have to be recalibrated.
Now looking at the VDU monitor, you should see a crosshair in the center of the screen. Move the cursor to the cross-hairs. Have the participant look at the cursor and when the crosshairs are stable, hit the E key. (If you happen to hit the E key just as the participant blinks, you will have to recalibrate them; just hit the R key to exit calibration and begin again). Now the cross hairs should have moved to the left of center (the distance it moves is based on the number of degrees of visual angle you are using to calibrate). Move the cursor to the crosshairs, and hit E again. Repeat for three other points and again for the center of the screen. Hit R to exit calibration. Hit N to run. Test the calibration by having the participants follow the cursor around the screen with their eyes. If their indicated eye position is not near the cursor (at least in the regions of the screen relevant to the experiment) then the participant must be recalibrated. You may need to do more than just repeat the calibration procedure; you may need to adjust the headgear and/or the pupil and crosshair settings. This is the part of eye tracking that is more art than science. You’ll have to just play around with different adjustments and find what works. (Note that we’ve had different results depending on how closely the crosshairs representing center of the pupil and corneal reflection (on the Eye Monitor) are spaced when the participant is looking at the center of the screen. You might try adjusting the headgear so that the mirror is directly in front of the eye—bringing the crosshairs close together—or down below the eye—placing them farther apart.)

On the Mac:
Select "Continue" from the EPAL Central Control interface. This should begin your experiment. (Note: the experiment examples given here require the experimenter to hit the "continue" button after each trial. This is to ensure that the experimenter will have the opportunity to recalibrate a participant should it be necessary.)
At the end of the last block, it will tell you that you've completed 5 of 5 blocks and it will give you a meaningless score. Keep hitting okay and then wait until the Listener comes back with new information in it. It takes a little time for the Mac to finish the experiment and writing data. Just make sure you don't quit the program during this time. The data will be in Apps->Active->EPAL-XXXX->protocols->partial by participants name and block.

5. Replay Scans
The replaying of scans may be conducted on any other computer equipped with MCL and the appropriate files (the data and some lisp, described below).
The appropriate files for the analysis are on Quasar. To watch the replay scans, begin by firing up MCL.
1. First several files will need to be created.
a. A file similar to new-icon.lisp (which is on Quasar) will have to be created for your experiment. It contains much of the lisp code necessary to do any post processing on the eye data, such as auto-compensating for a poor calibration and calculating the fixation points.
b. A.system file will also need to be created to load the relevant EPAL components, see epal-icon.system for an example.
c. Also, the post processing regions must be set-up. These are the regions that fixations will be attributed to for any post processing analysis. This can be done later, as a default set have already been created. To adjust these regions, after loading the system file, on the menu bar, select EPAL-XXXX, then "PP Settings". From here you can create new post processing regions.
2. Open MCL in Classic and load the system file (a .system file must be in the registry folder) "(load-system :epal-xxxx)
3. Load the file new-icon.lisp. (or the similar file you have created)
4. If you are not using auto-compensation, type (setf *use-auto-compensation-p* nil) to turn it off and make the replays faster.
5. Go to the EPAL-XXXX menu at the top->Globals->Check/Set. Choose your stimuli/resource file and your POR-regions file.
6. Select EPAL-XXXX/Do it!/Replay Scans. This will bring up the Scan Replay manager.
7. Click "Add" to add a participant's data to the manager. (This file will have a .1 extension, such as Sub5.1 and will take a few seconds to load).
8. Highlight the participant under "Available Protocols..." and click "Inspect." Then highlight a flavor under "Protocol Screening" and highlight a "Chosen Stims..".
9. From the scan replay manager, a number of options can be selected/altered, such as the replay speed, showing the post processing info, etc.
10.Then hit "replay." The scan should replay (it could take a few seconds).

References
Byrne, M. D., Anderson, J. R., Douglass, S., & Matessa, M. (1999). Eye tracking the visual search of click-down menus. Human Factors in Computing Systems: Proceedings of CHI 99, 402-409. Reading, MA: Addison Wesley.

	People   Facilities   Research   Projects   Miscellany   Search   |   Home

Last modified 5.12.2005