15th International Congress of Phonetic Sciences (ICPhS-15) Barcelona, Spain August 3-9, 2003

ANN F₀ Modeling for Female-Voice Synthesis in Spanish: Restricted and Non-Restricted Domains

J. M. Montero (1), L. F. D'Haro (1), R. Córdoba (1), J. A. Vallejo (2), J. Gutiérrez-Arriola (1), J. M. Pardo (1)

(1) Universidad Politécnica de Madrid, Spain
(2) Universidad de Oviedo, Spain

In this paper, we describe the modeling of the F₀ curve of a female voice in several Restricted-Domains and in a general-domain, aimed at developing a Speech Synthesis System for Spanish. For modeling F₀, we have used Multi-Layer Perceptrons, based on our previous experience with a male voice.
   For isolated speech and for continuous speech, the use of specialized MLPs is always preferred. The main difference between restricted and non-restricted domains is the relevance of the number of the recording carrier sentence for predicting F₀ in a restricted domain. The most relevant predicting parameters are stress, the position in the intonation group and the type of the group.

Full Paper

Bibliographic reference.  Montero, J. M. / D'Haro, L. F. / Córdoba, R. / Vallejo, J. A. / Gutiérrez-Arriola, J. / Pardo, J. M. (2003): "ANN F₀ modeling for female-voice synthesis in Spanish: restricted and non-restricted domains", In ICPhS-15, 563-566.