Emergent maturation from stochastic optimization in vocal development

Clément Moulin-Frier (1) , Jules Brochard (1) , Freek Stulp (2, 1) , Pierre-Yves Oudeyer (1)

1 Flowers - Flowing Epigenetic Robots and Systems
2 RV - Robotique et Vision

The aim of this paper is to propose and computationally support an original hypothesis regarding the predominance of jaw movements in infant speech development. We capitalize on previous research on emergent maturations from a stochastic optimization process on the arm domain. This work has shown that a quite simple optimization process, allowing a progressive learning of reaching arm movements by the minimization of a cost function, can display an emergent sensorimotor maturation: without precoding it, such an algorithm naturally recruits in priority proximal joints due to their wider range of effects. Here, we adapt this work to the vocal domain, by the use of a vocal tract model, and show that a similar effect can occurs on particular conditions, recruiting the jaw in priority due to the wider range of auditory effects implied by moving this articulator.

Domaines: Informatique; Sciences cognitives

Clément Moulin-Frier, Jules Brochard, Freek Stulp, Pierre-Yves Oudeyer. Emergent maturation from stochastic optimization in vocal development. 2014. ⟨hal-01100042⟩