(Epoch) resampling about mne-python

[agramfort]

that would be great to have this but my gut feeling is that piece of code
should land into scipy.signal

have you looked at:

http://docs.scipy.org/doc/scipy/reference/generated/scipy.signal.resample.html

I would send an email on the scipy-devel mailing list if somebody has done
this already and if there is an interest in scipy.signal.

option 2 is to have a resample method that uses scipy.signal.resample or
upfirdn http://code.google.com/p/upfirdn/ if availble.

makes sense?

[larsoner]

Yeah, I started trying to use upfirdn, and it was not as straightforward as I had hoped. Maybe for now we can try the scipy.signal.resample, and look at some results. We can always add more methods later (e.g., method='upfirdn' or whatever). How do you think on this:

1. Make generic function in mne-python that takes in a numpy array, in sample rate, and out sample rate, and (for now) basically wraps to scipy's resample, returning the resampled array. This way we could eventually change what we use for resampling, or extend it if we want.

2. Make methods in epochs (and wherever else) that call this generic function.

If this sounds reasonable, what would you like the function to be called for (1) the general function and (2) in epochs specifically? And what mne-python file should I put the general function in?

[agramfort]

Yeah, I started trying to use upfirdn, and it was not as straightforward
as I had hoped. Maybe for now we can try the scipy.signal.resample, and look
at some results. We can always add more methods later (e.g.,
method='upfirdn' or whatever).

makes sense

How do you think on this:

1. Make generic function in mne-python that takes in a numpy array, in
   sample rate, and out sample rate, and (for now) basically wraps to scipy's
   resample, returning the resampled array. This way we could eventually change
   what we use for resampling, or extend it if we want.

yes

2. Make methods in epochs (and wherever else) that call this generic
   function.

+1

If this sounds reasonable, what would you like the function to be called
for (1) the general function and (2) in epochs specifically? And what
mne-python file should I put the general function in?

I would say all called resample.

mne.time_frequency.resample()
Epochs.resample()
Evoked.resample()

I am relunctant to add a new signal module just for one function.

[larsoner]

Should I put the resample() function (the one that gets called by Epochs and Evoked, etc.) in filter.py? Seems like a natural fit.

[agramfort]

Yes

Le 26 sept. 2012 à 19:17, Eric89GXL [email protected] a écrit :

Should I put the resample() function in filter.py? Seems like a natural fit.

—
Reply to this email directly or view it on GitHub.

[larsoner]

So I've implemented a resample() function in filter.py. Comparing the results to MATLAB, it looks like I can actually get less (!) ringing in Python than I get using MATLAB's resample() (based on upfirdn) by using the 'flat' window, and zero-padding the data appropriately. I wrote it to operate on N-dimensional arrays along a given axis but haven't tested that yet, but I'm going to submit a pull request so people can try it out to see if they think it's reasonable.