[Eeglablist] synchronizing eye tracking and EEG data

[Olaf Dimigen]

Dear Merve,

 

it is hard to diagnose your synchronization problem without further information. I am also not sure what you mean by you “took out some saccades and fixations” from your raw eye-tracking (ET) data. Did you remove actual raw data samples? If you just added saccades or fixations via additional “message” lines (with correct timestamps), this should not be a problem. However, be very careful to stick to the existing raw data format, or the EYE-EEG importer won’t be able to parse the file anymore. 

 

Below some further comments that might be helpful:

 

How to avoid synchronization problems

------------------------------------------------------

I recommend to first synchronize the raw ET data with the EEG using EYE-EEG. You can then still detect the eye movement events in the merged ET/EEG dataset after synchronization, e.g. by using the velocity-based detection algorithm build into EYE-EEG. Detected saccades and fixations will be stored in EEG.event and from there, you can easily kick out fixation events or add more information to them (e.g.: which ROI was fixated?). Please note that the most recent EYE-EEG version also automatically imports the information of all other “messages” sent to the eye tracker during the experiment (e.g. coding experimental conditions), together with their new timestamp in the synchronized recording. This is also helpful for recoding fixation events. With Eyelink trackers, you can directly import eye movement events together with the ET raw data.

 

Common causes of synchronization problems: 

------------------------------------------------------

1.	Pausing the EEG recording during the experiment, which provides the “master clock” for synchronization. In contrast, pausing the ET recording is no problem.
2.	Selective loss of the start- and end-event needed to identify shared triggers by EYE-EEG. This can happen, for example, if the first event of the type chosen as the start event (e.g. “255”) is actually a different instance of this trigger in both recordings (because trigger “255” was sometimes lost in one of the recordings). In this case, check your data and try using different triggers as start- and end-events that exist in both recordings.

 

Typical synchronization quality

------------------------------------------------------

My sync tests with systems from three ET manufacturers (SR Research, SMI, Tobii Pro) and three EEG manufacturers (BrainAmps, Biosemi, EGI) show that EYE-EEG synchronizes the data with a mean absolute alignment error of 0.3 to 0.8 ms for all combinations of hardware tested (computed via common trigger pulses alignment; at sampling rate of 500 Hz). For individual triggers, the sync error should rarely exceed 2 samples (or 4 ms). This means that if everything went right, the mode of the distribution in the synchronization control plot should indeed be centered at zero, with no triggers exceeding plusminus 3 samples. 

 

Assessing synchronization via ET/EOG cross-correlation:

------------------------------------------------------

The current EYE-EEG version includes a new function to quantify synchronization accuracy also based on the actual time series data rather than just shared events. Specifically, this function computes the cross-correlation function between the (horizontal) eye-track and the bipolar (horizontal) EOG signal. Since both signals reflect the rotation of the bulbus, the peak of the cross-correlation function between both signals should be centered at lag zero or very close to it. You find more infos on this in the EYE-EEG documentation at http://www2.hu-berlin.de/eyetracking-eeg/tutorial.html#tutorial6.5.

 

Hope this helps,

Olaf

 

--

Olaf Dimigen
AG Biologische Psychologie

Humboldt Universität zu Berlin

Rudower Chaussee 18, 12489 Berlin, Rm. 2‘202

Tel -49 30 2093-4849

http://olaf.dimigen.de

 

 

 

 

Von: keskinmer at itu.edu.tr [mailto:keskinmer at itu.edu.tr] 
Gesendet: Donnerstag, 4. Januar 2018 11:08
An: eeglablist at sccn.ucsd.edu
Betreff: [Eeglablist] synchronizing eye tracking and EEG data

 

Dear all,

I use EYEEEG plugin of EEGLAB and am having troubles when synchronizing eye tracking (ET) and EEG data.

Before I synchronize the ET data with the EEG recording through shared events, I first filter the raw ET data outside EEGLAB (with a script to remove saccades whose size is >20 deg and also remove the corresponding preceding saccade. The script also filters out the fixation durations other than 50-1000ms long, as well as the corresponding following fixation).

I use this filtered ET file for syncronization and here is what EEGLAB returns for synchronization quality:

Warning: For 8 of 11 EEG events (72.7%), no ET event of same name was found within plusminus 5 samples. Possibly, some events were not transmitted to the eye tracker?Synch. error is the latency difference between matching ET/EEG events after synchronization.

I tried it with several data. When I check synchronization plots, EEG and ET events sometimes aligned and sometimes not, but the mode of distribution is never 0 (quality of synch graph).

Is it because ET data is not continuous anymore because I took out some saccade and fixations? I tried to look for an answer online, but I couldn't find anything helpful. Can anyone help me where I can be wrong?

All the best,
Merve

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20180109/6eb63790/attachment.html>