Comparison of time-domain vocal tract termination models

Vocal tract acoustic behavior is most concisely and conveniently described in the frequency domain. Given the inherent nonlinearities in the voice oscillator, modeling is best performed in the time domain. Two time-domain models for the vocal tract are tested. The first uses a "wave reflection" description of vocal tract acoustics, coupled to a time-domain mouth reflection/transmission model is described by a difference equation relating mouth volume flow and right-and left-running wave pressures at the mouth. The second model uses a pole-fitting approach to derive a robust, causal, and stable digital filter description of the vocal tract. The two models are tested by exciting each with identical time domain volume flows, then computing the mouth spectrum. These spectra are then compared to an analytical solution for mouth pressure in the frequency domain. Then, both models are coupled to a low-order vocal fold aeroelastic-aeroacoustic model, and differences in vocal fold vibration are reported.