Behavior of an adaptation-type neuron to the connected vowel speech sound has been investigated using an electronic model of the auditory nervous system. Samples of the speech sound are /aija/, /auwa/ and /aja/ of the synthesized speech (Fig. 1) and /maiasa/ of the human speech. An essential function of the adaptation-type neuron is a remarkable temporal sharpening effect due to the delayed lateral inhibition (Figs. 2, 3, 4, 13), so that the response to the connected speech sound is in many cases segmented temporally at each valley of the envelope of the speech sound (Figs. 6, 12). Furthermore, the additional segmentations to the secondary and third formants in the response occur for the frequency change (Figs. 5, 6, 7, 10(I), 12) and amplitude change (Figs. 8, 9, 10(II)) following the formant transition. Owing to these functions the response tends to be devided into a transitional part and a steady part of the connected speech sound. Adaptation-type neuron also behaves a stronger spatial sharpening than the primary and secondary neurons of the sustained-type. Therefore, it is suggested that the connected speech sound which a continuous signal is converted to the spatio-temporally discrete signals in the auditory nervous system.
抄録全体を表示