[Eeglablist] using ica weights in Matlab

[Yang Zhang]

Hi Eriksen,
 The original data has been whitened (to decorrelate the channels) before be submitted into the ICA decomposition procedure.
 So the ICA activations should be given by  ICA_activations = wts * sph * data;
 And the reversed algorithm should be given by eeg_activations=pinv(wts*sph)*ICA_activations;

Hope this is helpful for you!
-YANG!


Yang Zhang
------------------------------------
Department of Psychology
Northeast Normal University
Chang Chun, Ji Lin, China
130024
zhangyang873 at yahoo.com.cn
zhangyang873 at gmail.com
-------------------------------------





On Jun 25, 2011, at 8:47 AM, Jeff Eriksen wrote:

> I am hoping to do some specialized processing in Matlab using the icawinv and icaweights, part of the the ALLEEG Matlab structure. First I tried to reconstruct the EEG with one component removed but it did not look quite right. Next I decided to leave all components in and see if I could construct all the component time courses and then use them to reconstruct the whole EEG epoch. This did not come out right either. To get the components, I do this:
> 
> comp_acts = ica_weights * eeg
> 
> Then to get the EEG back I do this:
> 
> eeg-recon = icawinv * comp_acts
> 
> But find that the reconstructed EEG does not quire match the original EEG, and is much larger in amplitude as well.
> 
> I must not be understanding fully what these matrices are.
> 
> Thanks,
> -Jeff
> 
> _______________________________________________
> 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/20110625/6b1e54c1/attachment.html>