Has anyone had problems of synchrony between an eye-tracking system and the stimulus delivery software sending log messages to the eye-tracking system logs? And if so, what the possible sources of the problem could be, how to avoid the problem, and how to deal with it?
We have an MR-compatible eye tracker from MR Technologies hooked onto Arrington Research's ViewPoint EyeTracker software on one PC. On a different but connected PC, Neurobehavioral Systems Presentation software controls stimulus delivery. I have Presentation communicate to the eye tracking software's logs when my video stimulus starts and ends because I was told it was more reliable to manually start the eye tracking system rather than trying to control it through commands triggered from Presentation. As far as I understand, the eye-tracking system logs a line of data every 33ms even when there's tracking loss. I expect that I should have the same number of eye-tracking data lines between my video log markers -- so if my videos are 33 fps, I assume I should have the same number of eye data points as frames for a given video -- is that correct?
However, the eye-tracking data corresponding to a video is on the order of up to 3 seconds (1-80 data points) longer than the video. For example, for a random video and according to the Presentation log files for some random 2 subjects:
video X: 102 frames (25fps) = 4.1 sec length of video (according to Presentation log files and video)
sub1: 182 lines eye data @ 30fps = 6 sec of data marked as recorded during the length of video
sub2: 166 lines eye data @ 30fps = 5.5 sec
I am very hesitant to assume that the "video start" log marker I had Presentation send to the eye-tracker system log files really corresponds to when the video started (and then take only as much eye data as video length, ignoring the "video end" log marker -- or could this be a safe assumption?
Thanks in advance for any help and explanations!
I have very little experience with any of the technologies you use, however I am a big fan of time stamps. If the clocks on the devices you use are in sync, and you can log the timing of events, it should make your problem easier. To make sure, one can use a sudden event as a "clapperboard" to make sure there is actual sync.
Dear Gina,
we used the Arrington ViewPoint software in our lab with the 220USB system.
Some thoughts why you might get different frame numbers:
1) It is not a constant rate camera, so actually it runs at around 220Hz but not at exactly 220Hz. That may be the same case for your camera (although there might be less variability in the 30Hz version).
2) I found the marker insertion in the ViewPoint software rather unreliable (we used the remote Ethernet connection and the markers where sometimes dropped or capped at the start/end even with a security margin of starting the eye tracker some hundred ms before the experiment).
In conclusion, if your video start/end markers do not coincide with the data requisition start/end markers in the ViewPoint protocol, you are safe to discard the rest of the eye data. If the they to coincide, then the data might probably be corrupted.
Anyway, keep in mind, that the variable rate of the eye tracker will not guarantee you a one-to-one correspondence between video frames and eye tracker samples (the important property is the "delta time" in the ViewPoint protocol which gives you the the time interval between successive samples; if you have suffered a rate change, there will be variability in these numbers).
Hope that helps, Greetings, David
To add some comments to David’s answer:
- Some laptops (connected to eye-trackers) deliver different recording rates depending on their power supply: quite stable and high rate when connected to a regular power supply, but reduce their performance for very short intervals when using the battery – even when manufacturers say that this won’t happen.
- The same might be an issue concerning Ethernet communication.
- Sometimes there are background programs running und consuming the CPU’s performance (every time other programs consume too much performance, recording programs skip single frames)
- Sometimes programs do not immediately start when you start them.
- Badges
- Science topic
- Similar topics
- Psychology
- Cognitive Science