MCD distance is not 0

[Accomlish]

Thank you for your contribution I tried to calculate the MCD between the two identical sounds, but got a non-zero value

[chenqi008]

Thank you for your contribution I tried to calculate the MCD between the two identical sounds, but got a non-zero value

I have the same problem. Do you solve it?

[A321400841]

Maybe you need this project https://github.com/MattShannon/mcd/blob/master/bin/get_mcd_dtw

[tobiascornille]

Maybe you need this project https://github.com/MattShannon/mcd/blob/master/bin/get_mcd_dtw

That library doesn't support .wav files it seems. Did @Accomlish or @SamuelBroughton ever find the root cause?

[tobiascornille]

I did find that using the euclidian distance for DTW reduces the difference, but it's still not 0...

[jack159n]

I think there's a problem on the code, min_cost_tot += np.mean(min_cost) According to Librosa's document, it should be revised to min_cost_tot += min_cost[-1][-1]