Acoustic analysis of the vocal tract from a 3D physiological articulatory model by finite-difference time-domain method

Abstract

A faithful 3D physiological articulatory model was constructed in the previous work to investigate the mechanism of speech production. The model consists of the tongue, hyoid bone, jaw, larynx complex, bilateral piriform fossae, and vocal tract wall that includes the hard palate, soft palate, pharyngeal wall. In this study, we generated a series of time-varying vocal tract shapes in simulating vowel-vowel (VV) sequence by the 3D articulatory model and analyzed their acoustic characteristics using the finite-difference time-domain (FDTD) method. Area functions (AF) showed that the vocal tract shapes vary smoothly frame by frame, while the transfer functions (TF) transit continuously from vowel to vowel for the first four formants. Acoustic roles of the piriform fossae were also examined. The piriform fossa generated two spectral dips around 4 kHz in transfer function. Pressure distribution patterns at first four formant frequencies for vowel /a/ were examined, which are consistent with previous studies. Experimental results showed that our physiological model can describe the details of the vocal tract.