0303878 - URE-Y 20010108 RIV ES eng C - Konferenční příspěvek (zahraniční konf.)
Tučková, Jana - Šebesta, Václav
Data mining approach for prosody modelling by ANN in text-to-speech synthesis.
Anaheim: IASTED/Acta Press, 2001. ISBN 0-88986-301-6. ISSN 1482-7913. In: Proceedings of the IASTED International Conference on Artificial Intelligence and Applications - AIA 2001. - (Hamza, M.), s. 161-166
[IASTED International Conference Artificial Intelligence and Applications. Marbella (ES), 04.09.2001-07.09.2001 (K)]
Grant CEP: GA ČR GV102/96/K087; GA AV ČR IAA2030801; GA AV ČR KSK1019101 Projekt 01/01:4014
Grant ostatní: EU COST(XE) OC 258.10
Výzkumný záměr: CEZ:AV0Z2067918
Klíčová slova: speech processing * speech synthesis * neural nets * data mining
Kód oboru RIV: BD - Teorie informace

The contribution describes the artifical neural network (ANN) approach for modeling of fundamental frequency and duration of speech in text-to-speech synthesis. Methods for knowledge extraction from speech data are investigated and the number of ANN parameters for improvement of the generalization ability of ANN is minimized. The GUHA method for the choice of the most important input parameters and the standard pruning process of ANN for the optimization of generalization ability is applied.
Trvalý link: http://hdl.handle.net/11104/0114062