FE Modeling of Human Vocal Tract Acoustics.

Part I: Production of Czech Vowels

Finite element (FE) models open up new possibilities in simulating and understanding production of human voice and speech. This study presents FE models of the human vocal tract for the Czech vowels /a:/, /e:/, /i:/, /o:/, /u:/ which were created from magnetic resonance images of an adult male subject. The high number of elements in the FE models was reduced by introducing acoustic superelements. A novel algorithm for parameterization of the FE models was derived to allow modifying and tuning the model shape according to prescribed acoustic characteristics (formants) of the vocal tract. The sound radiation losses at the lips and sound absorption at the vocal tract walls were taken into account. The frequency modal characteristics of the three-dimensional (3D) and one-dimensional (1D) vocal tract models were compared. The results showed that the 1D models can closely replicate the behavior of the 3D vocal tract up to 3000 Hz. At higher frequencies transverse modes occurred in the 3D model which were not present in the 1D model. The developed FE models can be useful for studying the influence of vocal tract alterations (e.g. such as cleft palate or tonsillectomy) on acoustic voice quality.