Hi ,recently,I compute time-frequency analysis of scout using brainstorm,the graph seems not normal,espailly in the low frequency,there is a long-range time of activation,it's impossible because this is an instant finger tapping task，so I want to ask if I can change some parameters to deal with this situation,I have tried changes of FWHM and central frequency,but didn't work very well.
Hello,
Please refer to the corresponding introduction tutorial:
http://neuroimage.usc.edu/brainstorm/Tutorials/TimeFrequency
The simulation part explains why the transitions you observe are so slow, and also how to modulate the Morlet wavelet parameters to change the balance between the time and the frequency resolution.
With this approach, you try to estimate the power of an oscillation at a specific frequency. For this, you need a few cycles of this oscillations. At 3Hz, a full cycle lasts 333ms, the time resolution cannot be better than this, no matter what estimation method you use.
The image you posted here looks good to me. At least it doesn’t show anything abnormal, except that your epochs are maybe too short (select the option “Hide edge effects” - unless you have already removed them as explained in the tutorials).
If you want to experiment a different approach to time-frequency decomposition, we recently added the FieldTrip multitaper as an option (process Frequency > FieldTrip: ft_mtmconvol, refer to the FieldTrip documentation for using this method).
Cheers,
Francois