Abstract:

The acoustic shape of vowels is determined by the cross-sectional areas along the vocal tract. These areas can be predicted from the midsagittal distance, given certain assumptions about the relation of that distance to the area [P. Mermelstein, J. Acoust. Soc. Am. 53, 1070--1082 (1973)]. The present study used midsagittal MRI images of the vocal tract during sustained vowel production to provide areas as input to an articulatory synthesizer. The acoustic consequences of two methods of measuring distance will be assessed: computing distance with and without the epiglottis and with and without the uvula. For some vowels, the uvula extended into the oral cavity to different degrees across repetitions, resulting in large within-vowel variability for that region. Similarly, distances in the region of the epiglottis depended on the degree to which the epiglottis is detached from the root of the tongue, though this tended to be more consistent across repetitions. Closer matches, both in terms of formant frequencies and perceptual similarity, will indicate which method best represents the effective area predicted by the midsagittal distance. [Work supported by NIH Grant DC-02717.]