ESCA Workshop on Prosody Lund, Sweden September 27-29, 1993

Generation and Evaluation of Rhythmic Patterns for Text-To-Speech Synthesis

P. Barbosa, Gérard Bailly

Institut de la Communication Parlée, U.R. A. CNRS n° 368, ENSERG/INPG -University Stendhal, Grenoble 1, France

This paper presents a characterization of durational contours based on phasing relations between noticeable acoustic events and an internal clock. We generate segmental durations in two stages : the duration of a rhythmic programming unit is computed according to a reference clock and then is distributed among its segmental constituents. A perception experiment evaluates the necessity of the rhythmic patterns found in analysis. A method for mastering speaker's speech rate is described and is analysed to propose some guidelines for the integration of the pause phenomenon into automatic generation.

Full Paper

Bibliographic reference.  Barbosa, P. / Bailly, Gérard (1993): "Generation and evaluation of rhythmic patterns for text-to-speech synthesis", In Prosody-1993, 66-69.