Question about MFCC

Hi,

If I’m reading the code correctly, the features are MFCC with first 26 cepstrum coefficients, without delta or delta-delta?

I was wondering if you have tried delta and delta-delta, and if so, how good was it? A vague answer would be helpful! I’m asking because I don’t have a lot of computational resources, so would like to know if I should really try by myself. (I’m currently using data ~300hrs)

Thanks!

Correct.

We have not.

Is there a reason your group didn’t try that? My impression is that input features are something people experiment with.

There’s only so many things we have time, and GPUs, to try out :slight_smile:

If you experiment with different features please let us know how it goes!

Fair enough, thank you!