Dynamic neural model for categorical perception of speech

Abstract

The phenomenon of categorical perception has played an enormous role in the theory of speech perception. One reason is that categorical perception is at the interfacebetween the analog sensory signal and the discrete andsimbolic nature of language. Another phenomenon involvedin speech perception is selective adaptation. Adaptationeffects appear with respect to the location of the categoricalboundary along a speech continuum. To reproduce andaccount the categorization of speech sounds within a voicingcontinuum we propose a dynamic neural model. We showthat this model is able to reproduce the tipical patternsobserved in experiments of categorical perception, histeresisand adaptation effects. We compare the model predictions toexperiments with subjects. From this study we conclude thatcategorical perception may be understood as resulting fromcompetition within a neural representation of sensoryinformation.