Modeling speech production using the Neural Engineering Framework

A neurobiologically plausible model of speech production is introduced here using the Neural Engineering Framework (NEF). This approach allows detailed modeling of temporal aspects of action selection and action execution in speech production at the level of single spiking neurons. A preliminary architecture of our NEF speech production model is introduced and discussed in the second part of this paper. The first part focuses on an articulatory-acoustic model, generating acoustic speech signals on the basis of articulatory geometries. Our approach uses a small set of functional articulatory control parameters. Motor planning is based on the concept of speech or vocal tract actions (Kröger et al. 2010, Cognitive Processing 11: 187-205). A 2D-geometrical model is used and the acoustic speech signal is calculated using a reflection-type line analog model.