An M-mass vocal-fold excitation model and its relation to source–tract interaction
Jun Huang, Stephen E. Levinson, Don Davis, Scott Slimon · The Journal of the Acoustical Society of America · 2001
In this work, we propose an M-mass model of the vocal-fold excitation for articulatory speech synthesis. The vocal fold in this model is composed of M coupled oscillators. Each oscillator consists of a mass, a spring stiffness, and a dumper. The adjacent masses are coupled by a spring stiffness. The dynamic response of the M-mass model is described by a set of second-order ordinary differential equations called the equations of the motion. This M-mass model is combined with the Navier–Stokes equations to provide numerical solutions in an iterative way. First, the glottal volume velocity in the current iteration is used to determine the aerodynamic force exerted on each mass. Then the glottal area is calculated by solving the equations of motions given the aerodynamic force. This glottal area information is used to compute the glottal volume velocity in next iteration. We use this M-mass model to compute the glottal excitation signals including volume velocity and particle velocity. Experimental results show that we can observe ripples in the positive glottal opening interval of the particle velocity and volume velocity, which is strong evidence of source–tract interaction during speech production.