Partners and International Organizations
(English)
|
A, B, CZ, DK, FIN, F, D, IRL, NL, N, P, SI, E, S, CH, GB
|
Abstract
(English)
|
In the final year of this project, the prosodic components of our speech synthesis systems for French and standard German were refined, and the foundations for a new generation of more expressive systems was laid. The principles underlying the transition from the first to the second generation of prosodic control of speech synthesis were developed in the course of discussions within COST 258 and were documented in the COST 258 Volume ('Improvements in Speech Synthesis', 2001, Wiley Ltd). In the first generation of systems we aimed to reproduce the neutral-sounding and somewhat monotonous speech obtained when standard text is read aloud (news bulletins, fiction passages, etc.). In the next generation of systems, pronunciation will be varied with respect to type of discourse (e.g., spontaneous speech, interactive speech, announcements, etc.). This involves an empirical examination of the three major prosodic parameters, i.e., timing, fundamental frequency and intensity, in a much wider set of speech recordings. As a result of this transition, part of our work performed under the COST 258 heading this year concerned the automatiosation of research procedures, in order to facilitate the examination of large data sets.
|
