Sixth European Conference on Speech Communication and Technology
Wishing to obtain a more natural quality of the synthesized speech and to eliminate the disadvantages of the previous text-to-speech (TTS) systems evolved in our institute (Software ITC Cluj-Napoca, Romania), we experimented and developed a new synthesis method that combines the advantages of time domain signal processing with the requirement of pitch and duration modification (required by intonation). This paper presents some theoretical considerations, signal processing and implementation aspects of this pitch alteration technique that was adapted for the new version of the ROMVOX TTS system.
Full Paper (PDF) Gnu-Zipped Postscript
Bibliographic reference. Ferencz, Attila / Nagy, István / Kovács, Tünde-Csilla / Ratiu, Teodora / Ferencz, Maria (1999): "On a hybrid time domain-LPC technique for prosody superimposing used for speech synthesis", In EUROSPEECH'99, 1831-1834.