Skip to content

Draft: Audio overhaul

Oliver Kirsebom requested to merge audio-overhaul into dev-3.0

Overhaul of the audio module.

Main objectives:

  • Cleanup and simplify code base to make code maintenance easier
    • Remove support for "stacked" audio objects
  • Standardization naming conventions with the rest of Ketos code base (#4)
  • Resolve situation around supported audio file formats (#16)
  • Improvements to waveform/spectrogram interfaces, esp. related to instantiating new objects (#2)
  • Replace existing Numpy/Scipy algorithms for spectrogram computation with new Tensorflow algorithms (!10 (merged))

To-do's:

  • Implement the following audio representations:
    • Waveform
    • MagSpectrogram
    • CQTSpectrogram
    • MelSpectrogram
  • Drop: PowSpectrogram, GammatoneFilterBank, AuralFeatures
  • Drop support for parsing of time/frequency values from strings with SI units
Edited by Oliver Kirsebom

Merge request reports