Hello,
Two questions about the noise matrix: (1) the empty-room recordings length of my data is about 34s, how to choose baseline when I compute the noise matrix from the recording? (2) the tutorial computed the noise matrix based on notched data. Is it necessary to preprocessed the noise recordings like the raw data recordings such as filtering?
(1) the empty-room recordings length of my data is about 34s, how to choose baseline when I compute the noise matrix from the recording?
Use as much data as you can, the estimation of the noise will only be better. So use the full 34s segment.
(2) the tutorial computed the noise matrix based on notched data. Is it necessary to preprocessed the noise recordings like the raw data recordings such as filtering?
Yes. As stated here:
https://neuroimage.usc.edu/brainstorm/Tutorials/NoiseCovariance#Noise_covariance
Thanks for your reply, Francois. I will pre-process the noise recording with same sampling rate and same frequency filters as applied in the raw data. Should I also remove bad channels and run ICA (to remove machine noise as I did in the pre-processing of raw data)?
Should I also remove bad channels and run ICA (to remove machine noise as I did in the pre-processing of raw data)?
No, this is taken care of later automatically
Many thanks !
Hi, Francois. Sorry to disturb you. I tried to compute the noise matrix based on file A(notched, resampled and filtered as the raw data). Judging from the image, the result was strange. I'm not sure whether it's correct.
I also computed the noise matrix based on file B(notched, resampled as the raw data). This time, it looks pretty goood.
It looks indeed that there is something that doesn't work well.
Make sure this is not only a color mapping issue: if you have one abnormally high value somewhere, it might show only one yellow point somewhere and all the other values in blue.
Make sure you don't have bad segments covering the entire recordings, and that not all the channels are marked as bad. This computation is something relatively easy to understand, and it should give enough feedback for you to understand why there was no data in input.
Thanks, Francois. I have checked my data and I'm sure that there is nothing wrong with color mapping. No channel was not removed in the noise recording yet. Also, no bad segments.
Sorry, I mean no channel was removed in the noise recording yet.