I am currently using the Applied Science Laboratories (ASL) E6000 Eye-Head-Integrated System with a camera that should be able to collect eye movement data at 120, 240, and 360 frames per second. I am having a hard time, however, collecting data at these speeds because of issues with capturing both pupil and cornea reflections. Subjects experience discomfort when wearing the head gear for more than 20-30 minutes and inadvertent movement of the the hot mirror/monocle occurs. We are aware of several systems that claim to have the same accuracy as the ASL system (for example SmartEye, SMI's RED250, Tobi), but I would like to hear from your personal experience with such systems.
We use a desktop-mounted Eyelink 1,000 system (SR Research Ltd., Canada) with a temporal resolution of 1,000 Hz and a spatial resolution of 0.01° for our recordings. Participant’s heads are stabilized with a chin rest. This works very well, with healthy subjects as well as stroke patients.
Example
Article Non-lateralized auditory input enhances averaged vectors in ...
Example2
Article Antisaccade performance in Korsakoff patients reveals defici...
In my experience, commercial video eye tracking systems currently on the market should not be trusted for movements smaller then 0.5 degrees. The manufacturer's specs refer to theoretical performance or model eye performance.
For those with a binocular system, do this simple test: fixate for 30 seconds while recording both eyes. Then calculate the standard deviation of the difference between led and right eye vertical positions (vertical vergence). If this value is more than 5 arc minutes, then that is your tracker's noise level.
We use EyeLink 1000 and Tobii 1750 eye tracking systems. EyeLink 1000 accuracy is very good even when wearing a variety of glasses (includung stereo glasses). Tobii 1750 accuracy is not great and we don't use it when we are examining specific small object areas.
I don't know how EyeLink will perform in a dynamic high speed tasks as I have only used it for non moving serial object presentation.
The EyeSeeCam system developed by Erich Schneider's group in Munich is working well for us in studying saccades, pursuit and the VOR at about 220Hz.
I used the same setting as Tanja (Eyelink 1000, Desktop Mounted, 1000 Hz, monocular). But there is indeed a huge difference between the theoretical accuracy reported by the manufacturer (0.01 deg) and the 'practical' one I measured. To quickly estimate the 'practical' accuracy, I computed the average and max (RMS) deviation between dots appearing one at a time (at different positions on the display) and the recorded eye positions. The dots and eye positions were in fact collected for calibration/validation purposes before running an experiment. So, for 26 adult observers (13 dots per observer), I got an average RMS error of 0.43 deg. of visual angle, with the largest error being 1 deg. In attachment, you will find a histogram of the data. Left panel if for the average error per participant, and right panel is for the maximal error per participant.
HTH
We have the new SMI RED 500, which is the same as the RED 250 you mentioned, but with a sampling rate of 500 Hz. Seems to be one of the fastest remote eye-trackers currently available. I work with head-injured patients, and for this population this remote system is in my view one of the best choices. Advantages: remote control, so you can calibrate and record directly from your stimulus PC (Psychtoolbox, e-prime...); very stable tracking independent of lateral head movements up to 20 cm (the system has binocular tracking, but you can close one eye and it continues tracking with good accuracy); fine for long recordings. The remote control is particularly interesting (and easy to use) when you plan gaze-contingent displays. Disadvantages: 0.5 deg are in my point of view the lower limit of precision for a remote system, and can only be obtained when using a head-rest. Also, I admit I had some problems with setting up the remote control (the stimulus PC failed to establish an ethernet-connection), but SMI fixed the problem with an update within approx. 3 weeks.
I've had good experiences using the Eyelink Remote (head sticker, but no head gear) with both preschool and adult age groups. However, I'm not certain I would trust data from that eyetracker (or any remote eyetracker) down to .5 degrees. Are you testing a population that can't use a comfortable headrest?
Thanks for all the responses so far. Some of you asked about subjects and application of my eye tracking research. I conduct research in aviation, air traffic control to be more precise. Air traffic controllers in the en route environment work in teams of 2 or 3 and may move around a little bit as you may expect from an operator that works with a square, 29-inch display wears a headset for voice communications, and performs a lot of keyboard entries.
In our previous research using an ASL E4000 system at 60 fpm I had found that fixations on aircraft representation were much longer than those found in reading. A large part of the aircraft representation consists of alphanumeric data showing name, assigned and current altitude, and ground speed. The remainder shows heading, position, position history. My assumption has been that air traffic controllers have trained their visual system (unsupervised) to balance longer fixations to absorb the data and store it rapidly in long term working memory with saccades. The 60 fpm sampling rate may have masked micro saccades that would indicative of shorter fixations and simply reading the information rather than automatic acquisition of all the information at once. With the faster fpm I would like to either confirm or disprove my assumtions.
Hello Ben,
In my personal experience I got a very hard time with the ASL long range optic L6-HS in a MEG setting. We then changed for the long range EyeLink and it was piece of cake
- Badges
- Science topic