Natural sounds are complex phenomena which contain a mixture of events localized in time and frequency. As such, a statistical learning algorithm for synthesizing random instances of sound textures from an existing natural sound example is presented. It describes sound textures as a set of repeating structural elements subject to some randomness in their time appearance and relative ordering but preserving certain essential temporal coherence and across-scale localization.
Bibliographical noteFunding Information:
We presented a preliminary version of this article at the International Computer Music Conference (ICMC 99). This research was supported in part by the Israel Science Foundation founded by the Israel Academy of Sciences and Humanities. Ran El-Yaniv is a Marcella S. Geltman Memorial Academic Lecturer.