[Eeglablist] low-freq resolution issues in newcrossf
Makoto Miyakoshi
mataothefifth at yahoo.co.jp
Wed Apr 21 00:30:03 PDT 2010
Dear Nick,
You mean you are using the wavelet analysis, right?
I give you an example of the calculation.
If I'm wrong, someone please correct me (I'm not a specialist).
Suppose you are interested in 4Hz activity.
One cycle of 4Hz is 250ms (1000ms/4Hz).
If you use a 3 cycle wavelet, the length is 250ms*3cycles=750ms.
It means that your 750ms EEG data produces 1 pixel in the TF plot.
Thus, if your data length is 750ms, you obtain 1 time point.
If your data length is 755ms (suppose your EEG sampling is >=200Hz and the
time resolution ie. one time point of the TF plot is 5ms), you obtain 2
time points.
If your data length is 3000ms, you obtain (3000ms-750ms)/5ms=450 time
points.
If you want to obtain 200 time points and you have 3000ms data length, your
time resolution is (3000ms-750ms)/200=11.25ms. If you want 5ms time
resolution and 200 time points, then 5ms*200+750ms=1750ms data length is
required.
I hope now you understand calculating 0.06Hz is hard.
The lowest frequency of interest determines the window size, it should be
carefully determined considering the time-constant of your target EEG
component and (if event-related design) trial duration.
Makoto
--- Nick Bedo <nickbedo at yahoo.com> wrote:
> Hi all,
>
> I'm having an issue with some vertical streaks in my cross-coherence
> spectrogram at the lower frequencies (<10hz). My problem is similar
> to an archived problem
> (http://sccn.ucsd.edu/pipermail/eeglablist/2007/002006.html), but
> instead of having issues at high frequencies, I'm getting
> perturbations at the lower frequencies.
>
> Here is some of the output to give you an idea of my parameters:
> Adjust min freq. to 0.06 Hz to match FFT output frequencies
> Adjust max freq. to 49.99 Hz to match FFT output frequencies
> Using hanning FFT tapering
> Generating 1000 time points (128.1 to 1871.9 ms)
> Finding closest points for time variable
> Time values for time/freq decomposition is not perfectly uniformly
> distributed
> The window size used is 256 samples (256 ms) wide.
> Estimating 819 linear-spaced frequencies from 0.1 Hz to 50.0 Hz.
>
> Additionally, in an effort to increase overall resolution, I've been
> tweaking the default settings on newcrossf():
> DEFAULT_NWIN= 1000;
> DEFAULT_VARWIN= 0;
> DEFAULT_OVERSMP= 64;
>
> I have tried many minor tweaks to fix the issue, but nothing seems to
> be working. Any tips are appreciated!
>
> Thanks,
> Nick
>
>
> > _______________________________________________
> Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
> 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