View Single Post
Old 19th June 2020
Lives for gear
David Rick's Avatar
Under the hood

Originally Posted by Wolf LeProducer View Post
What tool do I need to split audio into a tone and noise part? For example, if I wanted to split a kick into its harmonic tone at the fundamental and the its upper mid rage, high rage click part?

Steinberg Backbone does this --- however, no way this is new tech.

What else does this?
The underlying DSP algorithm is called a "Singular Value Decomposition". It operates on a finite-length record, so it requires multiple frames (similar to FFT-based processing) and a "trajectory-tracking algorithm" that stitches the identified tones together across frames. The tonal components are those that come back with larger "eigenvalues"; everything with small values can be considered noise. Doing the cross-frame tracking and possibly associating some tones as overtones of other ones is part of the "special sauce" that each plug-in vendor has to invent for themselves.

If you've ever encountered the statistical analysis method called Principal Component Analysis, it's pretty much the same thing as a single frame of the processing described above.

David L. Rick