ibllib.io.extractors.training_audio

ibllib.io.extractors.training_audio.welchogram(fs, wav, nswin=262144, overlap=131072.0, nperseg=512)

Computes a spectrogram on a very large audio file.

Parameters:
  • fs – sampling frequency (Hz)
  • wav – wav signal (vector or memmap)
  • nswin – n samples of the sliding window
  • overlap – n samples of the overlap between windows
  • nperseg – n samples for the computation of the spectrogram
Returns:

tscale, fscale, downsampled_spectrogram