This can be useful in a transfer learning context for shifting frequencies and time scales into the pretrained domain. Eg whale sounds > speed up to look like birds on spec, or bat sounds > slow down to look like birds.
Speed change can be achieved by misrepresenting the audio sample rate. Audio is then resampled to expected sr.
There is a question of whether speed up / slowdown is sufficient or you need pitch shift or pitch invariant speed up/slowdown.
This can be useful in a transfer learning context for shifting frequencies and time scales into the pretrained domain. Eg whale sounds > speed up to look like birds on spec, or bat sounds > slow down to look like birds.
Speed change can be achieved by misrepresenting the audio sample rate. Audio is then resampled to expected sr.
There is a question of whether speed up / slowdown is sufficient or you need pitch shift or pitch invariant speed up/slowdown.