Enhanced area functions for noise source modeling in the vocal tract

The synthesis of natural-sounding fricatives based on a sourcefilter model is still a major challenge. While the filter is effectively modeled in terms of the vocal tract area function, noise sources are more difficult to model. Source properties critically depend on aerodynamic conditions and vocal tract geometry near the constriction in a way that is not fully understood. Therefore, noise source models usually assume different relations between noise source parameters and the aerodynamic state for different places of articulation. However, the place of articulation cannot be reliably determined from a dynamically changing area function, as in articulatory synthesis. Here we introduce the concept of an enhanced area function that adds the identity of the articulator that confines the vocal tract at the anterior-inferior side as a new layer of information to the classic area function. This allows to distinguish places of articulation and is therefore an effective representation of the vocal tract not only in terms of filter function but also for noise source modeling. A noise source model on this basis is presented.