[Eeglablist] Memory Issues: Importing Neuroscan .cnt files

[David Contreras Ros]

Hi Zach,

The problem stems from the fact that eeglab loads everything into memory 
before handling it. Most EEG analysis programs keep the information in 
the hard drive and load small parts of it when they need them, "on the 
fly". So, for instance, a 400MB file with 16-bit precision will turn 
into a 800MB matrix once loaded into EEGLAB and converted to float 
precision. That is only the raw data. If you perform ICA over that 
record, then you'll have another 800MB with the ICA-decomposed record, 
all loaded into memory. Whenever you perform operations on the data, 
eeglab will have to allocate some more memory to store intermediate results.

Probably the best solution is load the stuff in chunks, segment it and 
save it. Then concatenate the sets. You can write scripts for that. 
Every eeglab function is accessible from the command line in matlab, and 
the data model is accessible too. Check the workspace once you have 
started up eeglab and loaded a small dataset to get an idea of how 
eeglab organizes the data.

David.

Zachary Moran wrote:
> Hi all,
>  
> I'm part of a group at UCLA trying to implement EEGLAB to analyze 
> rather large, Neuroscan .cnt files collected with Scan 4.3.  The files 
> are 16-bit sampled at 1000 Hz, with an average size of about 400 MB.  
> I'm using a Dell/2.4GHz Intel with 1GB RAM running Windows XP 
> Professional and Matlab R2007a.  However, in trying to import 
> continuous files into EEGLAB, I get the following error:
>  
>
> EEGLAB error in function loadcnt() at line 381:
>
>  
>
> Error using ==> fread
>
> Out of memory.  Type HELP MEMORY for your options.
>
>  
>
> I've tried the usual memory saving options listed under "help memory" 
> to no avail.  I'm already running at 1GB of RAM, but does anyone know 
> whether the addition of more RAM would solve this problem or if there 
> are alternate solutions that have been found?
>
>  
>
> Any help or information would be greatly appreciated!
>
>  
>
> Thanks,
>
>  
>
> Zach Moran
>
> Graduate Student
>
> UCLA Lab for Clinical Affective Psychophysiology 
>
> ------------------------------------------------------------------------
>
> _______________________________________________
> 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
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20080126/c9ae0b44/attachment.html>