Statistical parameterization of f0 and formant estimation

Identify and formalize the underlying parameters that are estimated when computing fundamental frequency f0 and formant frequencies from speech recordings by developing a probabilistic model of speech signals in which f0 and formant frequencies are explicit, identifiable parameters amenable to statistical estimation and inference.

Background

Many phonetic studies rely on automatic or semi-automatic estimates of fundamental frequency and formants, whose accuracy is often checked manually. The prevailing perspective largely comes from signal processing, and a rigorous statistical framework specifying what these estimators target is lacking.

The authors explicitly mark this as an open modeling question and call for a statistical formulation that makes f0 and formants explicit parameters within a probabilistic model of the speech signal, enabling principled estimation and uncertainty quantification.