Acoustic characteristics of two types of finite element (FE) models of male vocal tract are compared in the paper. The first FE model was constructed from the data published by Story et al. (1996) on vocal tract geometry in English vowels /a/ and /i/. The design of this simplified FE model of the vocal tract was very laborious, however. For the second model, geometry of vocal tract shape for Czech vowels was obtained from a native Czech speaker using the magnetic resonance imaging (MRI) technique. An automatic procedure was developed which enabled to transform the MRI data files into a second, physiologically more plausible FE model. Acoustic frequency-modal characteristics of both FE models are computed and compared with the data known from literature.

Index Terms. Acoustic modes and formant frequencies, finite element models, magnetic resonance imaging technique