Second ESCA/IEEE Workshop on Speech Synthesis

September 12-15, 1994
Mohonk Mountain House, New Paltz, NY, USA

Improving the Robustness of Text-to-Speech Synthesizers for Large Prosodic Variations

Olivier Boeffard (1), F. Violaro (2)

(1) France Telecom, CNET, France; (2) UNICAMP/FEE/DECOM, Brasil

This paper describes a hybrid harmonic-plus-noise analysis/synthesis system developed to evaluate the potentiality of such systems in the pitch synchronous, concatenation-based, text-to-speech synthesizers. The main motivation for this work was the need for a system that enables greater prosodic modifications than the current generation of the PSOLA type systems. This is a primordial requirement in order to obtain a more natural sounding, high quality, text-to-speech synthesizer.

Full Paper

Bibliographic reference.  Boeffard, Olivier / Violaro, F. (1994): "Improving the robustness of text-to-speech synthesizers for large prosodic variations", In SSW2-1994, 111-114.