Hi,
I was wondering if there was anyway to do the time frequency analysis like
getting the power and building the tfr plots using the hilbert transform
instead of the morelet wavelets.
Thank you,
Aditya Singh
Biomedical Engineer
UT Austin
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.nmr.mgh.harvard.edu/pipermail/mne_analysis/attachments/20170704/1b3f687d/attachment.html
The Raw class has a method "apply_hilbert()" that you can use, but
you'll have to take care of bandpassing, iterating, etc. to get it the
way you want it, something along the lines of the CSP example in the
documentation (http://martinos.org/mne/dev/auto_examples/decoding/plot_
decoding_csp_timefreq.html#sphx-glr-auto-examples-decoding-plot-
decoding-csp-timefreq-py). The only "difficult" part is if you want the
phase instead of the envelope/amplitude, in which case you'll have to
do the phase transform yourself. (Maybe it's worthwhile to add a phase
option to the method?)
But done appropriately, you get the same answers with Hilbert, Fourier
or wavelets (Bruns, J Neuroscience Methods 2004).
Phillip
Just to clarify a bit - you can use the Hilbert transform for a TFR
representation by doing:
1. Bandpass filter your data (e.g., raw.filter)
2. Call apply_hilbert
1. If envelope=False, the output will be complex, so you can grab
either the phase or amplitude if you want
2. If envelope=True, the output will be a float and represents the
envelope of the signal (e.g. the ?power? of a TFR)
?