Formation of an auditory map for invariant perception of vowel sounds - listening to a variety of speakers to make unified vowel representation

摘要

We propose a neural network model that function as a cognitive map for the perception of vowel sounds. By simulating the model, we investigated neuronal bases for the encoding and perception of vowel sounds. The first network, which is tonotopically organized, detects spectral peaks called formant frequencies. The second network receives input from the first network in a convergent manner and detects the combinatory information of the first (Fl) and second (F2) formant frequencies. We trained the model with five Japanese vowels spoken by different people and modified synaptic connection strengths of the second network according to the Hebbian learning rule. We suggest that perception of unknown vowel sounds is possible if the sounds activate the neurons that are members of the cell assembly whose activation encodes the information about the category of the vowel to which the sound belongs.