Modeling speech processing using NENGO: First steps

Authors: Bernd J. Kröger

Abstract:

In our previous modeling of speech processing, SOM and GSOM approaches were used for building up a language specific syllable based speech action repository (SAR) within a neurobiologically plausible model of speech acquisition and speech processing. Because time as well as internal neural noise are not explicitly modeled in SOM and GSOM approaches, it will be difficult to simulate speech disorders in such a model. Therefore we now started using the NENGO neural simulator (http://nengo.ca/) in order to increase the neurobiological realism of our approach. It will be shown that this new approach allows modeling temporal aspects of action selection and action execution. Two examples will be given: (1) sequencing of syllable production and (2) switching between perceiving and reproducing words. Moreover the architecture of our NENGO speech processing model will be introduced and discussed in this paper.