search

Inria-NERV / happyFeat

A framework for simplying BCI workflow in clinical applications
https://happyfeat.readthedocs.io/
BSD 3-Clause "New" or "Revised" License
12 stars 4 forks source link

[bugfix] Review how frequency resolution is propagated in the app #9

Closed AsteroidShrub closed 4 weeks ago

AsteroidShrub commented 1 month ago

When using fres != 1 (ex: FFT size 1024 and fsamp 500) :

A lot of issues appear in the Visualization and Training parts. The frequency "factor" is either not used or user incorrectly (ex: the right frequency to choose could be 12Hz, but the bin should be 24, and the GUI and mechanisms randomly use 12 and 24...)

A big clean-up is needed on that aspect! ⚠️

AsteroidShrub commented 1 month ago

Linked to #3

AsteroidShrub commented 1 month ago

Another problem in the same topic:

"set frequency range" in AutoFeat also needs to be scaled. Users should need to work with actual Hertz, and HappyFeat should translate that to the index, considering scaling if necessary.

Ex: if nfft = 1024 and fsamp = 500, when using [7:35], HappyFeat understands (approx.) [3.5Hz : 17.5Hz]. This should not be the case, and HappyFeat should use indices [14:70] instead!

AsteroidShrub commented 4 weeks ago

Fixed by commit e43f1aa18606231827f4f2d45260f074e866c04c