Kenneth N. Stevens
Res. Lab. of Electron. and Dept. of Electr. Eng. and Computer Sci., MIT, Cambridge, MA 02139
In recent years, there have been significant improvements in models for predicting the properties of the radiated sound when the time-varying configurations of the vocal tract and laryngeal structures are known. These advances have been especially evident for models of consonant production, for which there are rapid movements of the articulators and consequent rapid changes in the properties of the sound. Experience gained with these models has made it possible to infer articulatory and laryngeal configurations and movements from appropriate measurements of the sound. Several examples are given to illustrate how these inferences can be made. These include: (1) detailed measurements of the changing spectra of frication and aspiration noise at the release of a stop consonant to infer the position and rate of release of the major articulator and the changes in the laryngeal configuration; (2) measurements to indicate the time course of the velopharyngeal opening; and (3) measurements from which the glottal configuration can be estimated. [Research supported in part by the NIH.]