Uncertainties in mechanical models of larynx and vocal tract for voice production

We have reviewed a number of studies of voiced sound articulatory synthesis, but all of them have been conducted using deterministic models. However, real voices are random processes. This paper is concerned with the discussion of uncertainties and random variables involved in the process of voice production, based on the maximum entropy principle, and using the mathematical-mechanical model proposed by Ishizaka and Flanagan. We discuss an implicit numeric method for synthesizing voice, which allows us to solve the associated inverse dynamics problem; that is, to identify the parameters of the model, given a specific voice signal. We can, then, analyze the sensitivity and random characteristics of the frequencies involved in the process, due to uncertainties in the parameters.