audio

audio copied to clipboard

While torchaudio provides a Mel-scaled spectrogram transformation (torchaudio.transforms.MEL), there’re a few additional spectral feature transformations that are extremely useful for pre-processing and data augmentation. For example, two feature transformations that I’d love to see in torchaudio are

• [x] Mel-frequency cepstral coefficients (MFCC) -- #83
• [ ] Chroma Energy Normalized (CENS) chromagrams.

I might take a stab at MFCCs after my finals (approx a month from now) if you fancy waiting that long :)

Anyone still looking into this? I would be interested in looking into chroma features if MFCCs are being taken care of 😄

bjuncek, could you confirm that you are working on the MFCC? I would be willing to work on this if your plans have changed. Thanks

@MuzzleTheWolf I'm working on it but the progress is slowed down by graduation requirements. If you can get it done before June, it's all yours.

Hey did this get anywhere? I'll look into implementing too if not.

I dont think anyone made progress. PRs definitely welcome :)

https://github.com/pytorch/audio/pull/83

I am interested in using chroma features within torch audio. Any progress on this? It seems that this issue has been open for a long time without any interaction. If possible, can anyone point to any alternative of pytorch based implementation of chroma?