I am trying to use the function "extract time series" for concatenated EEG files (total duration between 300 and 600 seconds). For "computing sources [2018] I have used the following settings and I have calculated the noise covariance matrix of each EEG file before this computation:
Does anybody have an idea what I've done wrong? Below you can find the general "subject structure" that I currently have (I've ran compute sources twice since I could not select "extract scout time series" the first time:
I've seem to get it working. I've unchecked the "Concatenate output in one unique matrix" box in the "extract scout time series" function. This was the problem.
"Concatenate output in one unique matrix" box in the "extract scout time series" function
This option saves all the extracted scouts time series as different signals in the same file. Therefore it requires all the signals to have the same length, which was not the case.
Another problem was that you had selected the concatenated file as well. Computing scouts for this file requires the reconstruction of the full source matrix (number of sources x number of time points = 13000x162000 = something that doesn't fit in memory).
I have selected the concatenated file without the "concatenate output in one unique matrix" box checked and it ran without a problem (+-2min for each concatenated file). Are my results then incorrect and is there an easy way to check this?
If your OS accepts allocating matrices of 16Gb, good for you!
You just need to be aware of the possible limitations of manipulating such gigantic files (and first of all, understanding clearly why this files are so big)
