Third International Conference on Spoken Language Processing (ICSLP 94)

Yokohama, Japan
September 18-22, 1994

A Scheme for Chinese Speech Synthesis by Rule Based on Pitch-Synchronous Multi-Pulse Excitation LP Method

Changfu Wang (1), Wenshen Yue (1), Keikichi Hirose (2), Hiroya Fujisaki (3)

(1) Department of Electronic Engineering, University of Science and Technology of China, Anhui, P. R. China
(2) Department of Electronic Engineering, University of Tokyo, Tokyo, Japan
(3) Department of Applied Electronics, Science University of Tokyo Noda, Japan

A pitch-synchronous analysis and synthesis method has been developed for Chinese speech synthesis by rule. This method is based on the multi-pulse excitation linear predictive (MELP) analysis. The wavelet transform is adopted to detect the instant of glottal closure (GCI) from the speech signal. Since the analysis is carried out pitch-synchronously within the interval of a pitch period bounded by two adjacent GCI's for a voiced speech segment, the major excitation occurring at a GCI is excluded from the analysis interval. Thus higher accuracy is achieved for vocal tract and excitation source parameters than that obtainable by conventional fixed frame methods. Since, in the synthesis process, the speech segments of one pitch period each are used as the units of synthesis for voiced speech, the pitch-synchronous scheme also simplifies the process of control parameter generation and increases the flexibility and controllability of the synthesizer.

Full Paper

Bibliographic reference.  Wang, Changfu / Yue, Wenshen / Hirose, Keikichi / Fujisaki, Hiroya (1994): "A scheme for Chinese speech synthesis by rule based on pitch-synchronous multi-pulse excitation LP method", In ICSLP-1994, 1679-1682.