@InProceedings{Horák2004_520,
author = {Petr Horák},
booktitle = {Studientexte zur Sprachkommunikation: Elektronische Sprachsignalverarbeitung 2004},
title = {Rule Based Sounds Duration Model for the Czech TTS System},
year = {2004},
editor = {Klaus Fellbaum},
month = mar,
pages = {191--193},
publisher = {TUDpress, Dresden},
abstract = {A phoneme duration model is a standard part of current text-to-speech (TTS) synthesizers. Our contemporary TTS systems have been using the relative duration modeling of separate speech units (diphones or triphones). The information about sounds borders is not available during speech synthesis. Only the information about borders of speech units is available. This paper deals with extension of the relative duration based TTS system. The aim is the TTS system with absolute duration modeling of separate sounds and the possibility of the using MROLA compatible Czech voices in the Epos TTS system. This task consists of extension of speech inventory about sounds margins information and extension of TTS system about separate sounds duration modeling. The main motivation for this work lay in the poor timing of our Czech triphone synthesis with relative duration modeling.},
isbn = {978-3-937672-65-6},
issn = {0940-6832},
keywords = {Sprachsynthese},
url = {https://www.essv.de/pdf/pdf/2004_191_193.pdf},
}