Title: Efficient exploration of articulatory dimensions
Authors: Paul Konstantin Krug, Peter Birkholz, Branislav Gerazov, Daniel Rudolph Van Niekerk, Anqi Xu, Yi Xu
Abstract:
The key to a successful simulation of speech acquisition with a parametric
articulatory synthesizer lies, inter alia, in the successful exploration of its articulatory
dimensions. However, such an exploration (regardless of the respective
algorithm) may be non-trivial due to the high dimensionality of a modeled vocal
tract and the associated high probability of creating unnatural or humanly impossible
vocal tract shapes. In this work, a method based on principal component
analysis is used to reduce the scope of motor space of the articulatory synthesizer
VOCALTRACTLAB. It is shown that such a technique can be used to increase the
computational efficiency of vocal learning simulations and thus may help to establish
better exploration-based acoustic-to-articulatory-inversion models.
Year: 2022
In session: Articulatory Synthesis
Pages: 51 to 58