Sixth International Conference on Spoken Language Processing
(ICSLP 2000)

Beijing, China
October 16-20, 2000

Restricted-Domain Female-Voice Synthesis in Spanish: From Database Design to ANN Prosodic Modeling

Juan Manuel Montero (1), Ricardo de Córdoba (1), José A. Vallejo (1), Juana Gutiérrez-Arriola (1), Emilia Enríquez (2), Juan Manuel Pardo (1)

(1) Grupo de Tecnología del Habla. Dpto. de Ingeniería Electrónica. Universidad Politécnica de Madrid
(2) Grupo de Tecnología del Habla-Departamento de Lengua Española-Universidad Nacional de EducaciDistancia-Ciudad Universitaria, Madrid, Spain
(2) Grupo de Tecnología del Habla-Departamento de Lengua Española-Universidad Nacional de EducaciDistancia-Ciudad Universitaria, Madrid, Spain Distancia-Ciudad Universitaria, Madrid, Spain

In this paper, we describe the development of a female voice in a Restricted-Domain Speech Synthesis System for Spanish. For the design of the database, we have used a greedyalgorithm approach that focus not only on covering a set of target phonemes, but also on mimicking the histogram of prosodic features from a larger database. For modeling the prosody, both duration and F0, we have used two Multi-Layer Perceptrons, based on our previous experience in unrestricted-domain modeling. The error normalised by the deviation is always below 0.7.


Full Paper

Bibliographic reference.  Montero, Juan Manuel / Córdoba, Ricardo de / Vallejo, José A. / Gutiérrez-Arriola, Juana / Enríquez, Emilia / Pardo, Juan Manuel (2000): "Restricted-domain female-voice synthesis in Spanish: from database design to ANN prosodic modeling", In ICSLP-2000, vol.1, 621-624.