Back to the main page.

Bug 257 - ft_freqanalysis_wltconvol, fourier not valid?

Reported 2010-12-01 23:41:00 +0100
Modified 2010-12-08 11:15:01 +0100
Product: FieldTrip
Component: core
Version: unspecified
Hardware: Macintosh
Operating System: Mac OS
Importance: P1 normal
Assigned to: Roemer van der Meij
Depends on:
See also:

Matt Mollison - 2010-12-01 23:41:16 +0100

Using fieldtrip from 20101130, if I use ft_freqanalysis with: cfg_freq.method = 'wltconvol'; cfg_freq.output = 'fourier'; ft_freqanalysis_wltconvol errors on line 104 because the variable csdflg doesn't get set in the prior if statement. It seems that ft_freqanalysis_wltconvol is only set up for pow and powandcsd. I'm still learning about this stuff. Has fourier just not been implemented in ft_freqanalysis_wltconvol, or will fourier never be a valid output for wltconvol?

Roemer van der Meij - 2010-12-08 11:14:33 +0100

Hi Matt, Fourier as output has not been implemented yet for this version of wltconvol yet. We are in the process of switching all freqanalysis functions to a new set of low-level functions, wltconvol is next on the list. When wltconvol is switched, fourier will be possible as output. Mathematically there is no reason why any method for frequency analysis, be it Hilbert-transform, wavelet-convolution or sliding time-window FFT, should not allow computation of a csd, power spectrum or keeping of the fourier-coefficients. Bruns (2004) has an excellent paper in JNeuroscience Methods showing the mathematical equivalence of these 3 methods if you're interested (given the proper parameters). Best, Roemer