AN AUTOMATED EVENT ANALYSIS SYSTEM WITH COMPOSITIONAL APPLICATIONS
Nick Collins · 2005
A modular system for event analysis is described which utilises psychoacoustically motivated onset detection to segment a musical audio signal. The target events have duration above the grain level in the 100-500mS range. Captured events are further analysed for features of pitch, integrated loudness and perceptual attack time. As a pragmatic approach to polyphonic audio, heuristics are specified to select and reject events meeting certain criteria based in statistical moments of instantaneous loudness designed to eradicate double hits and other unbalanced sound events. This system has been applied in both non-realtime composition within the MATLAB environment, and in a real-time form for interactive music via extension UGens and classes for SuperCollider.