Growing Sound Synthesizers using Evolutionary Methods

An approach to automate the design of sound synthesis algorithms (SSA) is proposed. It uses custom descriptions of the SSA in the form of cyclic topology graphs and acyclic tree graphs with ordered branches. Evolutionary methods are used in two stages: for suggesting topological arrangements for the SSA functional elements, and for optimizing the internal parameters of these functional elements. An analytical distance metrics that uses the complex spectrum of the target and test signals is proposed. An implementation of the approach is described and results are outlined.