Problems to detect ERP in EEG Signals

HHNStudentsHHNStudents Germany
in Research
Hello together,

two students and I have been working with the OpenBCI Mark VI Headset since April 2018. In our project at the University of Heilbronn we try to detect finger movements. To recognize the movement of the forefinger we use an external trigger on digital input D17. So, we see in the recorded data the input change from 0 to 1 if the finger pushes the trigger button. 
In Psychotoolbox for Matlab, we prepared a presentation that shows the test person arrows to the left or right. Depending on the arrow the test person pushes the corresponding trigger and hold it for 2 seconds. This happens for the left forefinger 119 times and for the right forefinger 121 times.
Our impedance of the electrodes C3 and C4 (10-20-System) are 30k ohm. We only look at these two channels with a sample rate of 250 Hz.
In preprocessing we filter the signal with a bandpass of 4 to 45 Hz. Then we search for changes of the digital input from 0 to 1. Then we cut the signals into pieces of the same length.
Now we look at each trail and the mean value of all trails. We expected to see ERPs on the signal, but it seems we only see random noise.

image


The picture shows the mean value of the movement of the right finger. We discussed the results with our professor, who has a lot of experience with EEG signals. She said the results are not good enough to detect movement. Her first assumption was that the impedance of the electrodes is too high. In this forum, however, it is written that an impedance of about 50k ohm is sufficient.
So we're a little confused why our signals look so bad and we can not read any information from the EEG signal.
Has anyone ever done a similar project?
Does anyone have tips and advice for us, what's wrong with our approach?

Thank you for your time.

best regards
Julian

Comments

  • wjcroftwjcroft Mount Shasta, CA
    edited June 2019
    Julian, hi.

    See this post and link,



    Just as a comparison, have you tried your ERP measurement with a 'wet' electrode system, such as paste or gel with cup electrodes?

    Here is Jeremy Frey's ERP measurements, comparing the Cyton with a g.tec amp,


    Regards,

    William

  • HHNStudentsHHNStudents Germany
    edited November 2018
    Hallo
    William,

     

    thank you for
    the answer.

     

    We tried
    the gold cup electrodes last week in two measurements. We used one electrode at
    the position of C3 and other one at C4. BIAS and SRB was connected to the left/
    right ear.

    Again, we like
    to detect the movement of the left and right forefinger. This time we only
    repeat each movement 50 times and calculate the mean over all movements of each
    kind.

     

    Our
    impedance of the electrodes C3 and C4 are 66k/60k ohm.


    https://picload.org/view/dclodwpc/leftfinger.jpg.html


    Unfortunately,
    the results do not differ in quality.

     

    So, the
    question remains why our signals look so bad and we can’t read any information
    from the EEG signal.

     

    best
    regards

    Julian



  • wjcroftwjcroft Mount Shasta, CA
    It sounds like something is odd with your signal averaging that would remove the noise,


    Jeremy was successful in his OpenBCI ERP measurements (above link). Can you try running your ERP calculations on some known valid EEG containing P300?
  • wjcroftwjcroft Mount Shasta, CA
    I believe in Jeremy's paper, he used an EEG cap, with injectable gel. See where he was placing his electrodes / reference. The 66Kohm you quote seems kind of high for a wet electrode connection. Seems like you could get lower than that with the right electrode setup. Try swapping around your electrode leads, you may have one going bad.
Sign In or Register to comment.

© Vanilla Keystone Theme 2026