[Eeglablist] Resampling "artefacts" 512-500 Hz

Downey, Ryan J RDowney at bme.ufl.edu
Mon Jul 8 07:29:34 PDT 2019


I just wanted to give a brief comment about low pass filtering to avoid aliasing before suggesting an alternative to resampling your EEG data.  

Low pass filters do in fact prevent aliasing but that is mostly only true for low pass filters applied during the data collection (analog circuits built into the amplifiers). As soon as you digitally sample the data, the aliasing has already occurred and it is too late for low pass filtering to be effective (high frequencies are already falsely appearing as low frequencies so using a filter to get rid of high frequencies doesn't help). So for example if you recorded at 1000 Hz sampling frequency and there was 4000 Hz noise that wasn't analog filtered at the time of collection, that 4000 Hz noise signal will persist in your digitally sampled data as low frequencies. Usually manufacturers have a low pass filter applied during data collection already, whether you are aware of it or not. Someimtes you can manipulate the parameters and sometimes not. The cut off frequency is typically set to 1/4 the sampling frequency). Note that no filters perfectly (100%) reject data in a specified frequency range so some aliasing may still occur but usually not much. The frequencies closest to the cut off are the ones most likely to slip by.     

The only case I can think of where applying a digital low pass filter (i.e. processing your data after it is recorded) would help remove aliasing is in the case of down sampling. For example if your raw data is recorded at 1000 Hz, it contains frequencies up to 500 Hz and if you want to down sample to say 200 Hz, then all the frequency content from 100 Hz (1/2 of 200) to 500 Hz (1/2 of 1000) needs to be removed BEFORE resampling. Thankfully EEGLAB automatically does this for you when you ask it to resample. Just remember that the low pass filter it applies can't really remove aliasing that already exists on your raw data. Technically speaking you could strongly filter the data to remove it, but you would have to know the frequency at which the aliasing is appearing and that would have to not also coincide with a frequency you are interested in (e.g. alpha in your data).  

Note that resampling the EEG signal from 512 to 500 Hz so it matches with your eye tracking 500 Hz data may not be the best approach. Even if you were able to resample your EEG data to 500 Hz without these artifacts, rarely do systems actually record at the sampling frequency they claim. For example, even if your EEG recorded at 500 Hz and your eye tracking recorded at 500 Hz, there is no guarantee they will line up. One system might actually record at 500.01 Hz and another at 499.99 Hz. If you just assume they both recorded at 500 Hz perfectly, then there will be a drift in their alignment over time; the longer your recording is, the worse the alignment will get. Ideally, you should have a start and a stop trigger signal or event markers common to your EEG and eye tracking system so that you can record these events in their local time systems and then properly align both the beginning and the ending points of the two systems together. Also, if you don't really need to have the eye tracking data and EEG data perfectly synched together (i.e. see their data streams simultaneously) but only need to know eye event timing (e.g.  when someone looked at a target), then you can simply linearly relate the eye tracking time to the EEG time (based on known common start/stop triggers) and then find the closest EEG sample that lines up with an eye tracking event of interest. 

Ryan J. Downey
Postdoctoral Associate
Human Neuromechanics Laboratory
Biomedical Engineering Dept.
University of Florida

-----Original Message-----
From: eeglablist <eeglablist-bounces at sccn.ucsd.edu> On Behalf Of Andreas Widmann
Sent: Monday, July 8, 2019 6:59 AM
To: Ziereis, Annika Carina <annika.ziereis at uni-goettingen.de>
Cc: eeglablist at sccn.ucsd.edu
Subject: Re: [Eeglablist] Resampling "artefacts" 512-500 Hz

Hi Annika,

from the sampling frequency I guess these are Biosemi data? Did you DC-correct (subtract mean per channel) or high-pass filter the data (before or after re-sampling)? In case not, could you please try first whether this solves the problem? In case the problem persists, could you please provide a sample file and the code of the preceding pre-processing steps (from eegh in case you used GUI).


> Am 04.07.2019 um 18:06 schrieb Ziereis, Annika Carina <annika.ziereis at uni-goettingen.de>:
> Dear all,
> (I'd like to say beforehand that I am a total beginner in signal 
> processing... so if you have an idea why this happened I'd be very 
> happy if you'd consider it in your explanation)
> For the EEG recording we use a sampling rate of 512 Hz which we want to down-sample to 500 Hz (we do simultaneous eye-tracking which records with 500 Hz and we want to do a co-registration).
> During the pre-processing I looked at the power spectral density and found a power increase for the harmonics of 12 Hz (so peaks at 12, 24, 36 ... Hz) after resampling to 500 Hz. These peaks were not visible in the  power spectral density plot of the raw data.
> It was especially pronounced for participants with a higher alpha power  and not so much for those with low alpha power.
> I played with some resampling frequencies to check how it looks like for 499 Hz, 501 Hz 497 Hz... and 256 Hz (to compare to a integer ratio).
> The "artefacts" only occur  for 499, 500 and 501 Hz-showing each the 
> harmonics of the difference in Hz e.g. for 499 Hz (which would be 13Hz less than the original sampling rate) I get the harmonic peaks for 13, 26 and so on.
> It disappears for Frequencies that are not in the alpha range (e.g. 15 Hz, for resampling to 497 Hz).
> Then I read that one could do low-pass filtering to avoid aliasing effects (which I have to admit- I don' t understand very well). So I tried to low-pass filtering of 40 Hz  (I just used 40 Hz because we did so in another study) and do the resampling again. This resulted in the same harmonics and super strange peaks in the filtered out frequencies.
> I uploaded a pdf with pictures here (hopefully, it makes it more clear 
> what I mean): 
> https://urldefense.proofpoint.com/v2/url?u=https-3A__github.com_aziere
> is_eegquestions_blob_master_downsampling.pdf&d=DwIGaQ&c=sJ6xIWYx-zLMB3
> EPkvcnVg&r=Xqv3i1MFyNQV9dhuL1e9BtqA0SbF1UWmfPtBzxDT0xY&m=v0YIIJolrv-LJ
> LgewPhhXUmnldQBmDQsGG--GMeQZ7c&s=GO4wqI0qRWKBRbjE7byitpUIyyirXImggx_hQ
> ikppFE&e=
> Has anybody encountered this problem before?
> Is this artefact problematic?
> Thank you very much for your help!!
> Best
> Annika
> ---
> Annika Carina Ziereis
> Georg-Elias-Müller Institute for Psychology
> - Affective Neuroscience and Psychophysiology - University Göttingen 
> Goßlerstr. 14
> 37073 Göttingen
> Phone +49 551 39 20623
> Email: annika.ziereis at uni-goettingen.de
> _______________________________________________
> Eeglablist page: 
> https://urldefense.proofpoint.com/v2/url?u=http-3A__sccn.ucsd.edu_eegl
> ab_eeglabmail.html&d=DwIGaQ&c=sJ6xIWYx-zLMB3EPkvcnVg&r=Xqv3i1MFyNQV9dh
> uL1e9BtqA0SbF1UWmfPtBzxDT0xY&m=v0YIIJolrv-LJLgewPhhXUmnldQBmDQsGG--GMe
> QZ7c&s=4b-4tw26Fmqr1XzFkHnzbn_kyRPnAwz1qkjKPCoB1Ik&e=
> To unsubscribe, send an empty email to 
> eeglablist-unsubscribe at sccn.ucsd.edu
> For digest mode, send an email with the subject "set digest mime" to 
> eeglablist-request at sccn.ucsd.edu

Eeglablist page: https://urldefense.proofpoint.com/v2/url?u=http-3A__sccn.ucsd.edu_eeglab_eeglabmail.html&d=DwIGaQ&c=sJ6xIWYx-zLMB3EPkvcnVg&r=Xqv3i1MFyNQV9dhuL1e9BtqA0SbF1UWmfPtBzxDT0xY&m=v0YIIJolrv-LJLgewPhhXUmnldQBmDQsGG--GMeQZ7c&s=4b-4tw26Fmqr1XzFkHnzbn_kyRPnAwz1qkjKPCoB1Ik&e=
To unsubscribe, send an empty email to eeglablist-unsubscribe at sccn.ucsd.edu
For digest mode, send an email with the subject "set digest mime" to eeglablist-request at sccn.ucsd.edu

More information about the eeglablist mailing list