ARCHITECTURE OF THE MULTIVOICE TEXT-TO-SPEECH SYSTEM
Abstract
Architecture of the multimodal text to speech synthesis system based on the voice conversion framework was proposed. Such system could be tuned to the specific speaker without any costs losses on the training phase and based on one speaker base, having in TTS system. Structural scheme for this type of the speech synthesizer, with the description of the functionality of the main blocks were presented. Their specific characteristics are synergy approach to the architecture and text-independent mode in the training phase.
About the Authors
V. A. Zakharyeu
Белорусский государственный университет информатики и радиоэлектроники
Belarus
Belarus
A. A. Petrovsky
Белорусский государственный университет информатики и радиоэлектроники
Belarus
Belarus
References
1. Лобанов Б.М. Компьютерный синтез и клонирование речи. Минск, 2008.
2. Sundermann D. // ICASSP. 2006. P. 81-84.
3. Duxans B. // PUC. 2006. P. 171-175.
4. Анализаторы речевых и звуковых сигналов: методы, алгоритмы и практика. // Под ред. А.А. Петровского. Минск, 2009
5. Bourlard H. Introduction to Hidden Markov Models. Lauseane, 2010.
6. Stylianou Y. // Springer. 2007. P. 502-532.
7. Захарьев В.А, Петровский А.А. // Докл. БГУИР. 2013. № 1 (71). C. 39-45.
Review
For citations:
Zakharyeu V.A., Petrovsky A.A. ARCHITECTURE OF THE MULTIVOICE TEXT-TO-SPEECH SYSTEM. Doklady BGUIR. 2013;(7):57-63. (In Russ.)
Views: 323
Email this article 