The APEX model: From articulatory positions to sound

Johan Stark, Christine Ericsdotter, Björn Lindblom, Johan Sundberg

The Journal of the Acoustical Society of America · 1998 · 11 citations · 0 references

Concepts

Abstract

From input specifications for lips, tongue tip, tongue body, jaw opening and larynx height, APEX constructs an articulatory profile which is used to derive an estimate of the associated area function and the formant frequencies. The geometry of the APEX vocal tract is customized according to x-ray data from an individual speaker [P. Branderud et al., Fonetik 98, 168–171 (1998) and www.ling.su.se/fon/publications/]. The input measurements are contours sampled as x/y points for the shape of fixed structures such as the mandible, hard palate, and posterior pharyngeal wall, and certain selectively chosen configurations of articulators whose shape and position vary (e.g., lips, tongue tip, tongue body). Observed tongue contours of /i/, /u/ and /a/, and ‘‘rest’’ are used to provide reference contours for the rules of APEX. The general philosophy behind the procedure is to have APEX represent certain key configurations as faithfully as possible and then derive intermediate articulations by a set of physiologically motivated interpolation rules [B. Lindblom and J Sundberg, J. Acoust. Soc. Am. 50, 1166–1179 (1979)]. An evaluation of the articulatory and acoustic performance of APEX will be presented. [Work supported by the HSFR and RJ research councils of Sweden.]