Hi everyone ! In the process of checking how Osekit computs PSD with the welch function I came upon some bug and unchanged behavior regarding the handling of multichannel audio. (see following code section)
1) Confusion between overlap and hop
As one can see on line 330 of the code, the overlap parameter, then implemented to scipy welch() function, is defined as equal to the hop parameter of the ShortTimeFFT scipy class. However it is false since the relation between both parameters is actually
overlap = window - hop.
2) Unsolved multichannel handling
|
self.audio_data.get_value_calibrated()[ |
|
:, |
|
0, |
|
], # Only considers the 1rst channel |
It appears on line 339 that this method is still only considering the 1st audio channel of the given file. A behavior that should have been changed by the PR #445, but that has apparently not been updated here.
Hi everyone ! In the process of checking how Osekit computs PSD with the welch function I came upon some bug and unchanged behavior regarding the handling of multichannel audio. (see following code section)
1) Confusion between overlap and hop
OSEkit/src/osekit/core/spectro_data.py
Line 330 in 8047f27
As one can see on line 330 of the code, the overlap parameter, then implemented to scipy welch() function, is defined as equal to the hop parameter of the ShortTimeFFT scipy class. However it is false since the relation between both parameters is actually
overlap = window - hop.
2) Unsolved multichannel handling
OSEkit/src/osekit/core/spectro_data.py
Lines 336 to 339 in 8047f27
It appears on line 339 that this method is still only considering the 1st audio channel of the given file. A behavior that should have been changed by the PR #445, but that has apparently not been updated here.