URL:

files/proceedings/2010/75.pdf

Abstract:

Existing methods for sound texture synthesis are often con- cerned with the extension of a given recording, while keep- ing its overall properties and avoiding artefacts. However, they generally lack controllability of the resulting sound texture. After a review and classification of existing ap- proaches, we propose two methods of statistical modeling of the audio descriptors of texture recordings using his- tograms and Gaussian mixture models. The models can be interpolated to steer the evolution of the sound texture be- tween different target recordings (e.g. from light to heavy rain). Target descriptor values are stochastically drawn from the statistic models by inverse transform sampling to control corpus-based concatenative synthesis for the final sound generation, that can also be controlled interactively by navigation through the descriptor space. To better cover the target descriptor space, we expand the corpus by au- tomatically generating variants of the source sounds with transformations applied, and storing only the resulting de- scriptors and the transformation parameters in the corpus.