Speaker: Pascal Perrier (Gipsa-lab Grenoble)
Date: January 18, 2018
We have been working for the last 20 years on the development of 2D and the 3D biomechanical models of speech articulators in the aim to better understand (1) how speech movements are constrained, (2) which degrees of freedom speakers have to deal with the goals of speech production (3) how the Central Nervous system integrate these properties to improve speech motor control and (4) how speech perception may use implicit knowledge of the physical characteristics of the speech production apparatus to extract relevant information from the acoustic and articulatory signals. In this talk I will present different steps in the development of these models and explain what kind of knowledge we could infer from all the models from the simplest 2D ones to the most complex ones integrating different 3D non-linear models interacting which each another.
I will also present some more recent work aiming at implementing an optimal feedback motor control scheme assuming that speech motor control could use simplified representations of the speech apparatus in the brain to simulate without any delay acceptable predictions of sensory inputs in order to deal with inaccurate motor commands or external perturbations of the vocal. Using biomechanical models in this context raises computational issues that I’d like to address in conclusion.