Back to the main page.

Bug 1199 - demeaning in ft_specest_mtmconvol should be performed in time-resolved fashion

Reported 2011-11-29 13:30:00 +0100
Modified 2013-01-17 10:45:43 +0100
Product: FieldTrip
Component: specest
Version: unspecified
Hardware: All
Operating System: All
Importance: P3 enhancement
Assigned to: Roemer van der Meij
Depends on:
See also:

Martin Vinck - 2011-11-29 13:30:42 +0100

Current demeaning procedure demains the LFP signal per trial, but this still leaves a non-zero DC component for a limited time-range.

Martin Vinck - 2011-11-29 13:33:53 +0100

since the output scales linearly with the DC, one could do the mtmconvol on an input with ones and then use a moving average convolution and multiply the spectrum at any time-point by the mva. Probably the first step can be simplified since the spectrum should be known if input is a vector of ones.

Roemer van der Meij - 2011-11-29 17:30:08 +0100

Hmmm, an interesting thought! Adding JM and Robert to the list for a discussion.

Boris Reuderink - 2012-02-01 13:54:43 +0100

Dear Martin, I have asked Roemer and JM, and it appears it has been discussed during the fieldtrip meeting, and the decision was made not to make the change. Instead, a high-pass filter could be used. Best, Boris

Roemer van der Meij - 2013-01-17 10:45:43 +0100

bug closing time