| Résumé |
We propose a new statistical model of musical timbre that han- dles the different segments of the temporal envelope (attack, sus- tain and release) separately in order to account for their differ- ent spectral and temporal behaviors. The model is based on a reduced-dimensionality representation of the spectro-temporal en- velope. Temporal coefficients corresponding to the attack and re- lease segments are subjected to explicit trajectory modeling based on a non-stationary Gaussian Process. Coefficients correspond- ing to the sustain phase are modeled as a multivariate Gaussian. A compound similarity measure associated with the segmental model is proposed and successfully tested in instrument classification ex- periments. Apart from its use in a statistical framework, the mod- eling method allows intuitive and informative visualizations of the characteristics of musical timbre. |
