On the Derivation of Area Functions and Acoustic Spectra from Cinéradiographic Films of Speech

Abstract

The objective of this paper is to determine whether it is possible to derive a transformation, valid for a given talker, for determining area functions corresponding to any speechlike articulatory configuration from measurements on lateral cinéradiographic films of the vocal mechanism. A tentative set of rules for accomplishing this transformation has been established. The rules are specified largely in terms of a series of functions that relate the cross-sectional area at each of a number of points along the vocal tract to linear dimensions determined from the x-ray films and incorporate the effects of jaw and velum positions. The validity of area functions derived through use of these rules has been tested by comparing formant frequencies measured for the corresponding utterances with those calculated (by digital computer) from the area functions. Webster's horn equation is taken as the basis for calculating formant frequencies from area functions. The corresponding spectra are also computed for comparison with the measured spectra of the speech sounds, making certain assumptions regarding source spectra and locations and resonance bandwidths. In general, agreement between computed and measured formant frequencies by the two methods is within a few percent for the first three formants for a variety of consonant and vowel configurations. [Work supported in part by the U. S. Army Signal Corps, U. S. Air Force Office of Scientific Research, and U. S. Office of Naval Research; in part by the National Science Foundation (grant G-16526), the National Institutes of Health, U. S. Department of Health, Education, and Welfare (grants MH-04737-03 and NB-04332-01), and the National Aeronautics and Space Administration (grant NsG-496); and in part by the U. S. Air Force (ESD) under contract AF19 (628)-3325.]