David R. Williams
Sensimetrics Corp., 26 Lansdowne St., Cambridge, MA 02139
A series of simulations was conducted to determine the precision needed to specify orifice area parameters for synthesis of intervocalic voiceless obstruents. Time variation of glottal and oral cross-sectional areas was controlled using a set of quasi-articulatory parameters [K. N. Stevens and C. A. Bickley, J. Phon. 19, 161--174 (1991)]. Values of flows and pressures were computed using a low-frequency equivalent circuit model of vocal tract aerodynamics. Intraoral pressure values were mapped into KLSYN88 synthesizer source parameters via a set of equations. Rate of change in oral constriction area at closure/release was varied over a 10:1 range, as was the size of the narrowest fricative constriction. Maximum glottal opening was varied over a 3:1 range. Timing of the glottal movement relative to the oral movements was varied by (plus or minus)30 ms. While peak glottal and oral flows were significantly affected by the manipulations, the characteristic acoustic attributes of the two obstruent types were retained over a wide range of the magnitude variations. Obstruent voicing was also fairly robust over changes in relative timing, although extreme timing misalignments did not yield especially good exemplars of the category. [Work supported by NIMH.]