MR imaging of the vocal tract during vowel production

Modeling of vocal tract behavior during speech production requires accurate measurements of the entire vocal tract. To accomplish this, magnetic resonance (MR) imaging was performed on a 0.5-T system with a custom-made, double-loop, wrap-around coil. A midsagittal section of the region from the glottis to the lips was imaged during sustained production of five American English vowels. A gradient-spoiled gradient-echo sequence was used to achieve a 4-second imaging time. Images showed clear profiles of the entire vocal tract and exhibited essential articulatory features of vowel production. It was possible to measure midsagittal widths at various points along the vocal tract with 1-mm resolution. The measurements were used to construct a tube model for prediction of the acoustic output of the vocal tract. It is concluded that risk-free MR imaging promises to be an important means of acquiring data for vocal tract modeling, with applications to speech pathology, linguistics, and artificial speech.