Application of neural networks to duration modeling in a Spanish text-to-speech system

Abstract: We are going to show the application of neural networks in one of the critical modules of a text-to-speech system: duration modeling. Our objective is the accurate prediction of segmental duration. We present a complete environment in which to decide which are the most relevant parameters and how to code them. We will compare two systems: an unrestricted-domain database for a male voice and a restricted-domain environment for a female voice. The restricted-domain offers several advantages to the modeling: the variation in the different patterns is reduced, and so most of the decisions about the parameters are based in more significant results. The result is a very low prediction error, specially in the restricted-domain case