Vocal Tract Model Research Articles

Further Theoretical and Experimental Bases for Quantal Places of Articulation for Consonants

Previous theoretical studies of the acoustic behavior of a vocal-tract model consisting of a tube with a narrow constriction located at different points along its length have suggested that there is a series of constriction positions for which the sound output has well-defined quantal attributes. It has been hypothesized that these positions correspond to places of articulation that are used to produce consonants in a variety of languages. Recent work has refined the theoretical basis on which these places of articulation are predicted, and has shown that the shape of the constriction as well as its positions plays a role in determining the distinctive acoustic attributes of the output, particularly for consonants produced with the tongue blade. Furthermore, data on the acoustic properties of stop and fricative consonants from several different languages have been collected and have been shown to be consistent with the theoretical framework. The predicted places of articulation account for consonants in the pharyngeal, uvular, and velar regions, and several consonant classes in the dental-alveolar-palatal regions. [Work supported in part by National Institutes of Health and in part by the Office of Naval Research.]

Acoustical Consequences of Lip, Tongue, Jaw, and Larynx Movement

In the past, numerous attempts have been made to develop quantitative models of the human vocal tract. The present paper describes such a model and some of its acoustic properties. It has been constructed with a view towards finding a set of parameters that are physiologically “natural.” Thus it differs from earlier schemes in that the mandible has been incorporated as an independent parameter. Other novel features can be interpreted to represent larynx height, the rounding-spreading action of the lip musculature, and “genioglossal,” “styloglossal,” and “hyoglossal” components of tongue shape. The paper describes the acoustic effects of varying these parameters systematically. The model is capable of generating most of the vowel qualities known to occur in the languages of the world. In general, well-known vowel sounds are found along the extreme contours of the total acoustic space characteristic of the model. They are not necessarily located at acoustic “plateaus” or regions insensitive to articulatory imprecision. The model specifies the notion of “possible vowel articulation” in terms of parameters that are not in a one-one relation with distinctive feature dimensions such as open, close, high, low. A discussion of the implications of these results for universal phonetic and phonological theory concludes the paper. [This work was supported by a research grant from The National Institutes of Health.]

Articulatory Analysis of Speech by On-Line Computing

An on-line interactive computer program is described for calculating the parameters of an articulatory model of the human vocal tract corresponding to measured amplitude spectra of speech sounds. The computational procedure comprises two stages: (1) determination of the poles and zeroes of the amplitude spectrum; (2) calculation of a vocal-tract shape having the computed pole-zero pattern. In the first stage, a match to the input spectrum is synthesized from a predetermined number of resonance curves of individual pole or zero pairs having adjustable frequency and bandwidth values. In the second stage of the computation, the cross-sectional area of the vocal tract and the degree of nasal coupling are iteratively adjusted until a satisfactory match to the pole-zero pattern of the input spectrum is achieved. Specification of the vocal-tract shape in terms of the Fourier coefficients of the logarithmic area function leads to a rapidly converging matching strategy. Results for the analysis of vowel-consonant-vowel sequences are given.

Further Experiments Using a Larynx Model

Research has continued on the model of the larynx described at a previous meeting of the society. [T. H. Crystal, “Model of Larynx Activity during Phonation,” J. Acoust. Soc. Am. 33, 93(A) (1965)]. The model has been adjusted to exhibit the same trends in subglottal pressure, average flow rate, and vocal intensity measured by Isshiki in experiments on human subjects [N. Isshiki, “Regulatory Mechanism of Voice Intensity Variation,” J. Speech Hearing Res. 7, 17–29 (1964).] This is achieved by adjusting spring tension and rest position of the simple-harmonic oscillator that is used to model the mechanical facets of vocal-fold activity. Additional data have been gathered on the loading effect of simple vocal-tract models on the operation of the larynx model. The model also has been used to demonstrate the waveform changes that might be expected when a talker breathes helium-rich air. [Work supported in part by the National Institutes of Health, U. S. Department of Health. Education, and Welfare, by the U. S. Air Force Cambridge Research Laboratories, Office of Aerospace Research, by Project MAC, an MIT research program sponsored by the U. S. Department of Defense.]

Phonemic Feature and Vocal Feature : Synthesis of Speech Sounds, using an Acoustic Model of Vocal Tract

Phonemic Feature and Vocal Feature : Synthesis of Speech Sounds, using an Acoustic Model of Vocal Tract

Toward the Specification of Speech

This is an interim report on studies of the specification of speech sounds from acoustical measurements. Methods based upon analysis, synthesis, and vocal tract models are described. Included are the results of preliminary measurements on the vowel sounds of 25 speakers. Some of the problems in specifying the vowel sounds as indicated by these results are discussed.