The Phase Vocoder [1] was originally introduced as an alternative approach to speech coding but has won much greater acceptance among the music community as a tool both for sound analysis and composition [2]. Although dormant for some time, there has been a resurgence of interest in AM-FM speech signal descriptions in the last ten years [3], [4]. With the intention of building on some of the new ideas proffered, the aim of this work is to first consider their application to musical signals. It then demonstrates how paramaterisation of the extracted AM-FM information using EPS (Exponential Polynomial Signal) models allows modification of the individual large- and small-grain features of the AM and FM components, thus providing a new way for generating audio effects.

Download