ABSTRACT
In this paper we investigate the possibility of an intelligent system for sound synthesis that would allow composition with sounds thought of in terms of abstract descriptions and operations rather than low level computer programing. Our research work is aimed at looking for plausible guidelines and strategies to map a composer’s intuitive notion of sound structure to the parametric control of electronic sound synthesis. This paper shows how we attempted to approach the problem by means of a compilation of a few well known expert systems design techniques used in Artificial Intelligence (AI) research. We begin introducing the problem from a musician’s point of view. Then we focus on two major issues of AI/Music: knowledge representation and machine learning. We propose a suitable representation scheme for sound synthesis followed by an inductive learning experiment towards an example prototype synthesiser. We also introduce the prototype functioning and suggest further work.
