Signal processing for the analysis of musical sound

The signal environment of musical sound is a complex one with significant underlying structure. Past attempts at machine understanding of musical sound have been fraught with problems. We have been taking a new approach to the problem - using advanced signal processing algorithms as "sources of knowledge" in an Expert System developed to understand the sound. The emphasis in this paper is on the signal processing techniques. These include parametric modeling approaches to the segmentation of the sound, spectrum analysis, and harmonic set formation. We describe the problem specifics vis-à-vis signal processing and the techniques developed in progress to date to perform the analysis.