Sixth European Conference on Speech Communication and Technology

Budapest, Hungary
September 5-9, 1999

A Novel Model TD-PSPTP for Speech Synthesis

Yan Huang, Bo Xu

National Laboratory of Pattern Recognition, Institute of Automation, Chinese Academy of Sciences, Beijing, China

In this paper, a novel approach based on time-domain pitch-synchronous point-to-point (TD-PSPTP) model for speech synthesis is presented. Compared to TD-PSOLA, which is currently one of the most popular concatenation methods, TD-PSPTP model provides a wider range of pitch and time modification. The quality of synthesized speech by TD-PSPTP shows to be high, especially its capability of overcoming reverberation, existing in TD-PSOLA when there is a drastic prosodic modification. The computational expense of TD-PSPTP model is no higher than that of TD-PSOLA. It provides an efficient way for the real time implementation of synthesis system.

Full Paper (PDF)

Acoustic Example

Bibliographic reference.  Huang, Yan / Xu, Bo (1999): "A novel model TD-PSPTP for speech synthesis", In EUROSPEECH'99, 2303-2306.