Second ESCA/IEEE Workshop on Speech Synthesis

September 12-15, 1994
Mohonk Mountain House, New Paltz, NY, USA

A Strategy for Changing Speaking Styles in Text-to-Speech Systems

Masanobu Abe, Hideyuki Mizuno

NTT Human Interface Laboratories, Yokosuka, Japan

For enhancing the performance of text-to-speech(TTS) systems, this paper proposes the extraction of rules specific to particular speaking styles. This strategy makes it easy for a TTS system to synthesize speech in various speaking styles. As the first trial, three speaking styles were examined. Specific rules were generated for 1st and 3rd formant frequency, Fo height assignment for minor phrases, average phoneme duration, duration lengthening in a syllable followed by a pause or sentence end, and speech power gain. The rules were integrated into a conventional TTS system and listening tests confirmed the good performance of the proposed strategy.

Full Paper

Bibliographic reference.  Abe, Masanobu / Mizuno, Hideyuki (1994): "A strategy for changing speaking styles in text-to-speech systems", In SSW2-1994, 41-44.