Towards the Automatic Generation of Sound Synthesis Techniques: Preparatory Steps

The field of automatic generation of synthesized sound has been focused usually in the estimation of parameters for a predetermined synthesis technique (or topology) to try to reproduce a particular target sound (1), (2). While this is very plausible, it is known that different sound synthesis techniques perform better with different types of sounds. Therefore, some parameter matching techniques can give poor results when using a fixed topology. A search algorithm that includes the “sound synthesis techniques space” as part of the search can be used to select from a limited set of synthesis techniques (i.e. the “classic” synthesis techniques) before applying the parameter-matching algorithm. An even more adventurous goal would be to develop an algorithm capable of search through a broader set of synthesis techniques (i.e. classic and more). This kind of algorithm would be even able of suggesting “new” (i.e. not classic) synthesis techniques (topologies).