Physiological control of low-dimensional glottal models with applications to voice source parameter matching

This study explores the possibility for physical models of the glottis to accurately reproduce target glottal flow waveforms. A set of rules is proposed for controlling a two-mass physical model through activation levels of laryngeal muscles. The proposed rules convert muscle activities into physical quantities such as fold adduction, mass, thickness, depth, stiffness. Numerical simulations of the glottal model, controlled through this set of rules, are used in order to construct a codebook between muscular activations and a set of relevant acoustic parameters of the voice source, such as foundamental frequency, open quotient, speed quotient, and return quotient. The paper explores the potentialities of the derived codebook for the control of the glottal model, and its applications to voice source parameter matching. A fitting procedure is developed, and it is shown that the parameters of the physiologically-controlled two-mass model can be matched to fit target signals (glottal flow pulses) with good accuracy.