Feature Selection and Composition using PyOracle

Regions set according to feature IR. The feature with the highest IR value is considered to be the most informative of a given segment.

CREL researcher, Greg Surges, recently published a paper describing a system which uses the Audio Oracle al- gorithm for music analysis and machine improvisation.

PyOracle is a new machine improvisation and analysis system in the family of software built around the Factor Oracle and Audio Oracle algorithms (Assayag et al. 2006; Francois, Chew, and Thurmond 2007). PyOracle is the first software to use Audio Oracle (AO), a graph structure built using features derived from input audio, in the context of machine improvisation. AO is useful for both analysis and generative purposes, and will be described below.