Source separation and layering structure01 Aug 2016
Audio source separation is the isolation of sound producing sources in an audio scene (e.g. isolating a horn section in a big band).
Nonnegative Matrix Factorization (NMF) is a popular source separation method. It learns a dictionary of spectral templates from the audio. Separation via NMF needs external guidance to group spectral templates by source.
Composers often indicate important structural groupings by introducing instrumental groups one by one, as seen above in String Quartet No. 1, Op. 27, Mvt IV, Measures 1-5 by Edvard Grieg. We use these grouping cues to inform NMF. We also use NMF to find when new source layers enter.