audio-resampling-in-python icon indicating copy to clipboard operation
audio-resampling-in-python copied to clipboard

Update torchaudio resampling

Open carolineechen opened this issue 4 years ago • 1 comments

Resolves #6

Summary of Changes

  • used the latest version of torchaudio (0.9.0), which contains various resampling improvements
  • replaced torchaudio resample_waveform with the new functional.resample, since compliance.kaldi.resample is being deprecated
  • removed functional.resample from impulse graphs and replaced functional.resample spectrograms with transforms.Resample using a larger lowpass_filter_width parameter, similar the julius plots. functional.resample should numerically match transforms.Resample, so plotting both seems redundant
  • updated README to reflect resampling times in notebook as run on my local machine

Let me know if you have any questions or suggestions!

carolineechen avatar Jun 30 '21 20:06 carolineechen

Huh, I did a review very soon after you created this PR. Where has it gone?! Will have to do it again.

jonashaag avatar Aug 19 '21 20:08 jonashaag