strongest reference for SSVEP and alpha?

edited January 2016 in Research
Hi,
I'm having trouble getting meaningful EEG signals off MYSELF. I managed to get both ALPHA and SSVEP off my wife.
Setup:
32b OpenBCI
Headware from FRI (dry electrodes) and Velcro-based saline electrodes per @wjcroft.
Earclips for BIAS and SRB2 at two ears.
Fresh battery pack.
Dark, silent room.
Laptop on batteries.
All unnecessary electrical equipment turned off.
OpenBCI GUI used for raw recordings.
Custom Matlab code for post-processing (spectrograms, etc).
SSVEP with blinking movies (courtesy of Chip) and a blinking LED off an Arduino Uno board.

I was able to reproduce Chip's plots of his SSVEP experiments (from his raw data). This validates the post-processing SW.
I managed to get good ALPHA and SSVEP readings off my wife.
However, I'm failing to produce meaningful readings off myself for ALPHA and SSVEP experiments.

Here's a sample spectrogram of a session. http://dropcanvas.com/7ff34
The left-hand side upto the first vertical white dashed line is just me staring at the monitor with eyes OPEN.
The right-hand side from the first vertical white line upto the second white line is me sitting with eyes closed to evoke ALPHA waves.
I don't see them.

Furthermore, if you look at the plots, you will notice a relatively strong signal at 5, 7 and 9 Hz. These appear almost always at back electrodes no matter what I do.

Any ideas?

Thx,
David

Comments

  • I spent a lot of time with SSVEP and OpenBCI. I tried many configuration and I think that the best location for SRB electrode is Oz. I also couldn't achive SSVEP with SRB electrode mounted at ear lobe.  
  • nekrodezynfekator.
    With SRB @ the forehead I'm beginning to see some meaningful responses.
    Interestingly, the O1 and O2 responses are stronger than Oz.
    The responses decrease as I move up along the midline, i.e. POz is weaker and Pz is the weakest (almost nothing there).
    I managed to see responses both off video and LED blinking.
    Interestingly, sometimes I pick up frequencies half the stimulus frequency, i.e. I pick up 5 Hz off a supposedly 10 Hz stimulus.
    Since the stimulus is 100 msec ON 100 msec OFF, the overall frequency is 5 Hz.
    I recall @chipaudette experimented with this, but don't remember what his conclusion was.
  • U should check how to compute frequency.

    f=1/T 

    T-period
    f-frequency

    T=time ON + time OFF.

    So, i your case:

    T = 100 ms + 100 ms = 200 ms = 0,2s
    f = 1/0,2s = 5 Hz
    what u see is correct ;)
  • Except that other times I see 1/Ton frequency.
    His conclusion was that the brain responds to a changing signal within a certain frequency window, be that ON-ON or ON-OFF.
  • Well, finally after a lot of experimenting, I'm getting meaningful SSVEP results from myself.
    Thx for everybody who's commented / shared their experience.
    Best electrode position: POz
    2nd best: Oz (noisier than POz)
    SRB: Cz
    Bias: linked ears with earlobe clips
    Electrodes that didn't work well: O1, O2, Pz (O1 and O2 too noisy?, Pz too weak signal?)
    Electrodes: FRI, no gel, Velcro headware
    Amp: OpenBCI 32-bit, 8 channels, 250 Hz sampling rate
    Stimulus: 60 Hz monitor, sine-modulated luminosity at whole display, 10 sec ON, 5 sec black, 5 sec baseline w/ white screen
    Sequence: 4 12 5 13 6 14 7 15 8 16 9 17 10 18 11 19 20 Hz
    Stimulus that didn't work: single red LED w/ sine luminosity modulation from Arduino Uno (perhaps too weak a source?)
    50 Hz notch, 2-45 Hz BP (both IIR, had to employ zero phase non-casual arrangement)
    OpenBCI GUI used for raw data recording, Matlab for offline post-processing

    Here're the spectrograms of POz and Oz:
    image

    Next steps: 
    - will try ICA / BSS to see if I could denoise the signal better
    - will repeat for ASSR (and use Tx, FTx electrodes)
  • Could u give me the steps how to do such nice spectrograms? :)
  • I use spectrogram to calculate the power spectrum of electrode readings.
    You have to give spectrogram meaningful parameters (window, overlap, NFFT), otherwise spectrogram will be BS.
    This gives me t,f and ps.
    Then imagesc to plot 10*log10(ps).
    I look at the image and adjust Clims so that the image looks nice (didn't find good heuristic yet).
    I changed the OpenBCI FW to include user pressing the PROG button, which I use as markers for trial onsets and ends.
    I use discretize to partition t into bins for each trial.
    I plot vertical lines to delineate trials on the figure returned by imagesc (hold on).
    Then I use text to put annotation boxes with stimulus frequencies.
    xlabel, ylable for labels.
    colorbar for colorbar.
    xlabel on handle returned by colorbar to put a label on colorbar. 

    The rest is Matlab coding.

    Hope this helps,
    D
Sign In or Register to comment.