[Eeglablist] isolating a CNV with ICA

Ronald Phlypo ronald.phlypo at ugent.be
Fri Sep 11 00:56:36 PDT 2009


Dear Keith,


If it concerns slowly evolving potentials, or potentials with a slowly 
decreasing autocorrelation function, second-order (statistics) methods 
based on non-proportional autocorrelation functions might be more 
appropriate than the non-gaussian (higher order statistics) versions of 
ICA.

Some references:

    * *SOBI:* Adel Belouchrani and Karim Abed-meraim and J.F. Cardoso
      and E. Moulines, 'A Blind Source Separation Technique Using Second
      Order Statistics', IEEE Trans on Signal Processing, pp434--444
      (45), 1997
    * *TDSEP: *A. Ziehe & K-R Müller, 'TDSEP - an efficient algorithm
      for blind separation using time structure', Proc. of the 8th Int'l
      Conf on Artificial Neural Networks, pp675-680, 1998
    * *AMUSE: *Lang Tong and Ruey-Wen Liu and Soon, Victor C. and
      Yih-Fang Huang, 'Indeterminacy and Identifiability of Blind
      Identification', IEEE Trans. on Circuits and Systems, pp499-509
      (38) 1991
    * *but also* L. Molgedey and Schuster, H.G., 'Separation of a
      Mixture of Independent Signals using Time Delayed Correlations',
      Physical Review Letters, pp3634-3637, (72), 1994


SOBI is included in the EEGLab package in an implementation of 
Belouchrani and Cichocki [sobi.m] and might be worth the try either for 
separating the signal of interest, either for a non-parametric slow 
varying potential wave subtraction as a preprocessing step.


Good luck!


Ronald


Joseph Dien a écrit :
> You might try doing a PCA in the temporal domain instead of an ICA in  
> the spatial domain, if the time course is sufficiently consistent.
>
> Cheers!
>
> Joe
>
> On Sep 8, 2009, at 1:01 PM, Keith Yoder wrote:
>
>   
>> Hi all,
>> Our group has EEG time-series data collected with a BioSemi  
>> ActiveTwo 128-electrode cap.  Our paradigm includes a Go-NoGo, with  
>> an alerting stimulus (S1) preceding the presentation of the Go-NoGo  
>> stimulus (S2).  S1 and S2 are separated by 3 seconds.  When we  
>> examine ERPs from single electrodes, we observe a typical contingent  
>> negative variation (CNV).  However, when we run ICA (runica), none  
>> of the components isolate the CNV.  We've tried highpass filtering  
>> at 0.5hz (obviously to high), 0.01hz and without any highpass  
>> filtering, all without success.  Has anyone else run into a similar  
>> problem?  Does anyone have any additional advice for isolating a CNV  
>> with ICA?
>> Thanks,
>> Keith Yoder
>> --
>> Research Assistant
>> Belmonte Autism Lab
>> Cornell University
>> _______________________________________________
>> 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
>>     
>
>
> --------------------------------------------------------------------------------
>
> Joseph Dien,
> Senior Research Scientist
> Center for Advanced Study of Language
> University of Maryland
> 7005 52nd Avenue
> College Park, MD 20742-0025
>
> E-mail: jdien07 at mac.com
> Phone: 301-226-8848
> Fax: 301-226-8811
> http://homepage.mac.com/jdien07/
>
>
>
>
>
>
>
>
> _______________________________________________
> 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
>   
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20090911/820e193b/attachment.html>


More information about the eeglablist mailing list