FUNCTIONAL MODEL TO REPLICATE THE REPRESENTATIONS OF HUMAN VOWELS IN THE PRIMARY AUDITORY CORTEX

摘要

Previously we investigated representations of human vowels in the primary auditory cortex (Al) of awake cats [Qin, et al, J. Neurophysiol, 99, 2305-2319, 2008]. Herein we develop an auditory model with a multichannel neural pathway to replicate such representations in Al neurons. The model consists of seven blocks: the basilar membrane (BM), inner hair cell (IHC), primary auditory nerve (AN), ventral cochlear nucleus (VCN), inferior colliculus (IC), medial geniculate body (MGB), and Al. Weighted summations across the channels are introduced into the VCN, IC, MGB, and Al blocks because lateral inhibition is observed for these connections. We evaluated the constructed model by simulating the responses of Al neurons to human vowels. The output from the model of five Japanese vowels successfully represented the formants of vowels. Because this model generates pulse trains similar to neural firings in Al neurons, it can be used to convert acoustical signals into electrical stimuli as part of an auditory brain-machine interface for individuals with severe hearing impairment.