First European Conference on Speech Communication and Technology

Paris, France
September 27-29, 1989

Speech Synthesis Based on Vocal Tract Region Theory

Mohamed Mrayati (2), René Carré (1)

(1) Institut de la Communication Parlee, INPG-ENSERG, Université Stendhal, UA au CNRS, Grenoble (2) Scientific Studies and Research Center, Damascus

A new concept in speech production based on distinctive spatial regions along the vocal tract and distinctive modes, is used in speech synthesis with new insights. It is a step ahead towards articulatory speech synthesis. A 9-parameter model for speech synthesis is described (Region Synthesizer). These parameters are tightly correlated with articulatory parameters. This model has the merit of making complex Acoustic-Articulatory-Phonetic relations easier. The parameters of the model are pseudo-orthogonal. The model is capable of synthesizing vowels and consonants. A new command strategy refers to movements of eight regions vertically around the neutral axis of the vocal tract. It uses extensively synergy and compensatory movements with respect to the mid-point of the vocal tract. We will present in this article results on vowel, vowel-vowel synthesis and examples of coarticulation in VCV cases.

Full Paper

Bibliographic reference.  Mrayati, Mohamed / Carré, René (1989): "Speech synthesis based on vocal tract region theory", In EUROSPEECH-1989, 2172-2175.