Generating Source Streams for Extralinguistic Utterances

This paper demonstrates how self-organizing dynamic systems can be used to generate suitable source-streams for synthesizing extralinguistic utterances in the context of a source-filter voice synthesizer. Conventionally, the source-stream is simulated using two types of generators: one generator of white noise and one (or more) generator(s) of periodic pulses. Then, by carefully controlling the amount of signal that each generator sends to the filters, one can roughly emulate the behavior of the vocal folds whether or not they are tensioned. A number of variations have been proposed in order to furnish the source-stream with more realism, but to our knowledge none of these has addressed the needs of extralinguistic utterances. We propose that source-streams suitable for synthesizing extralinguistic utterances can be produced using granular synthesis techniques coupled with self-organizing dynamic systems. In this paper we present an example which uses a cellular automaton-controlled granular synthesizer. The paper begins with an introduction to the problem followed by a brief commentary on the limitations of the standard approaches. Then it details an example implementation of our proposed approach and concludes with a commentary on the results.