Question in CQT2010v2 and VQT

[VenkatSBitra]

I could see the option of "center=True/False" for CQT1992v2 but was unable to find a similar option for CQT2010v2 and VQT. I was wondering what these modules assume by default.

Also, I was wondering if there is a way for me to generate center frame times for each frame generated from CQT1992v2 with both center=True and center=False, CQT2010v2 and VQT.

[migperfer]

Hi VenkatSBitra,

thanks for this issue! Indeed there is no centering for CQT2010v2, nor for CQT2010.
Actually I just noticed that the ammount of padding is different depending on wether padding mode is "constant" vs "reflect", which sounds not right? (cc @KinWaiCheuk)

Seems like centering can be an improvement. If you want to implement it I suggest you to use librosa' s imlementation as guidance, also, feel free to make a contribution if you got it working :)

[KinWaiCheuk]

padding is different depending on wether padding mode is "constant" vs "reflect", which sounds not right?

You are right, they should have the same padding length.
But I remember for reflect, there is a case where we don't have enough sample to be reflected, and the padding ended up shorter.

[dorienh]

The center parameter is not currently exposed in CQT2010v2 or VQT, both default to non-centred framing. For frame time calculation you can derive centre times manually as t = (frame_index * hop_length) / sr, or t = (frame_index * hop_length + n_fft/2) / sr if you need the centred equivalent.

Note that VQT had a correctness bug (it didn't correctly reduce to CQT when gamma=0) that has been fixed in nnAudio2, the modernised successor: https://github.com/AMAAI-Lab/nnAudio2 — worth switching if you're doing active development.