What is an Epoch in MNE and Wavelet transformation

✉️ Mailing List Archive (read-only)
1

External Email - Use Caution

Hello,

I have three questions:

1- In order to apply Wavelet transformations, do we apply it on the whole
signal t=0...n, where n is the number of seconds in a recording? Or we do
something similar to STFT, where we window the signals (with some overlap),
and apply Fourier transformation on each epoch (epoch here makes sense to
me).

2- What statistical or non-statistical features that we should extract from
Wavelet coefficients (cA and cD)?

3- Does MNE Epochs does "windowing", or it does something else. I couldn't
get that from what it says in the documentation.

The problem I'm working on is Emotion (baseline + many seconds) using EEG.
So the signal for each trial (e.g. watching a video) is (n_channel,
seconds*sampling frequency).

Thanks,
Badar Almarri
Graduate Student
Dept. of Computer Science and Engr.
University of Connecticut
badar.almarri at uconn.edu
<http://www.linkedin.com/pub/badar-almarri/48/b09/8b8>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.nmr.mgh.harvard.edu/pipermail/mne_analysis/attachments/20180814/57060f56/attachment.html

2

External Email - Use Caution

The wavelet transformation in MNE implicitly windows the signal based on
the number of cycles per frequency (the Morlet kernel for each frequency is
only some finite number of samples long). In the end, the amplitude in each
frequency band is the result of a convolution with the Morlet kernel, so
the estimates are continuous, if I recall correctly.

As for what you extract from it, you get estimates of amplitude and phase
(from the complex values) at each time point. From there you have many
options depending on what you're actually trying to understand in your data.

Epochs are basically windows of time from the raw data. Take a look at the
epochs object tutorial and documentation, and try it out on your data.
Hopefully this will clear up any remaining confusion.

Eric
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.nmr.mgh.harvard.edu/pipermail/mne_analysis/attachments/20180822/7073ae9e/attachment.html