[Eeglablist] Peak shift problem

Toby Fernsler toby at sccn.ucsd.edu
Mon Jun 4 17:09:42 PDT 2007


Sometimes two ICs are very similar to each other, to the point that they 
are basically the same thing. For example you might get several ICs that 
clearly represent one eye movement. In this case you group them 
together, and treat them in the same manner. Infomax ICA isn't a perfect 
method, it doesn't always separate what we consider a single source into 
a single component, and it can miss sources. It is not intended for 
separating out frequency components, that's what frequency transforms 
are for. What it does well is identify spatially independent sources of 
large power.

Toby Fernsler

linhaisheng318 wrote:
> Hello,
> I am now using ICA to identify the components of mixed spectra. But 
> one sharp peak of some mixed spectra is a few wavenumber shifts of 
> that of ther other mixed spectra. So, when ICA is applied to identify 
> the components, there would be two ICs who are similar but only a few 
> wavenumber shifted.
> I do not how to deal with this situation. I have used some 
> preprocessing methods to deal with the shift problem. But it still did 
> not work. I wonder whether there are some efficient methods, no matter 
> preprocessing, change of ICA algorithm and so on, that can solve this 
> problem.
> Thank you very much for your help.
> Best wishes!
> LIN
>
>
> ------------------------------------------------------------------------
> ÈËɽ ÈË º£ Ê¢ ¾°£¬¾¡ ÔÚ ÃÎ »Ã Î÷ ÓÎ 
> <http://event.mail.163.com/chanel/xyq.htm?from=163_NO6>
> ------------------------------------------------------------------------
>
> _______________________________________________
> eeglablist mailing list eeglablist at sccn.ucsd.edu
> Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
> To unsubscribe, send an empty email to eeglablist-unsubscribe at sccn.ucsd.edu



More information about the eeglablist mailing list