Sixth International Conference on Spoken Language Processing
(ICSLP 2000)

Beijing, China
October 16-20, 2000

Articulatory Synthesis Using a Vocal-Tract Model of Variable Length

Zhenli Yu (1,2), Shangcui Zeng (1)

(1) Dept. of Information and Electronic Engineering, Zhejiang University, Hangzhou, China
(2) Motorola China Research Centre, Shanghai, China

A method of articulatory synthesis using a vocal-tract model with variable length is proposed. The vocal-tract length is derived prior to the unique determination of vocal-tract area parameters incorporated with a codebook that maps formants to vocal-tract length is used. A two-dimensional interpolation function for irregularly spaced data is conducted to confine vocal-tract length in the first and second formant plain to generate the codebook. The reflection-type line analog (RTLA) model of articulatory synthesis is employed to generate speech sound. Variable vocal-tract length yields the issue of discretetime implementation with multi-rate sampling of the RTLA model. Multi-rate sampling conversion is conducted.


Full Paper

Bibliographic reference.  Yu, Zhenli / Zeng, Shangcui (2000): "Articulatory synthesis using a vocal-tract model of variable length", In ICSLP-2000, vol.1, 693-696.