1999 Fiscal Year Final Research Report Summary
Physiological and neural network studies on phoneme analysis by the auditory cortex
| Project/Area Number |
10480068
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (B)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
Intelligent informatics
|
|---|
| Research Institution | Toyohashi University of Technology |
|---|
Principal Investigator |
HORIKAWA Junsei Toyohashi University of Tech., Dept. Knowledge-based information Eng., Prof., 工学部, 教授 (50114781)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
YAMAGUCHI Yoko Tokyo Denki Univ., Dept. Information Science, Prof., 理工学部, 教授 (00158122)
|
|---|
| Project Period (FY) |
1998 – 1999
|
| Keywords | Auditory cortex / Optical recording / Neural network model / Phoneme extraction / Auditory recognition / Wave sonagram hypothesis / lateral inhibition / Guinea pigs |
|---|
| Research Abstract |
Neural mechanism of auditory cortex for extraction of the auditory information has not yet known despite many electrophysiological, Psychological and information engineering studies. In this research, we have investigated a possible mechanism of the auditory cortex for feature extraction of phoneme by using an optical recording method and a nerd model of the auditory cortex that process spectro-temporal of sounds.
Optical recording with a 12×12-channel photodiode array and a voltage-sensitive dye (RH795) was made from the auditory cortex a guinea pigs anesthetized by Nembutal (30 mg/kg) and neuroleptanalgesic (doroperidol 0.3 mg/kg/h and pentazocine 1.1 mg/kg/h). Spatio-temporal responses to pure tones, frequency-modulated (FM) sounds and a complex sound with two FM components (CFM) presented to the ear contralateral to the recording cortex were studied. Pure ton stimulation at different frequencies clearly showed the isofrequency bands and the tonotopic organization in the primary audi
… More
tory cortex (AI). In response to FM sounds, the active spot moved obliquely crossing the isofrequency bands. When the direction of the frequency sweep was changed, the active spot moved in the opposite direction. The response to the CFM was complex : active spot to each FM component moved obliquely crossing the isofrequency bands with mutual inhibition.
Based on these results, we proposed a wave sonagram hypothesis and made a neural network model of the auditory cortex using 2-dimensionally distributed neural oscillators with excitatory and inhibitory connections. Computer simulation revealed that the model reconstructed the fundamental characteristics of the activity (wave) propagation observed in the auditory cortex. Furthermore, the simulation demonstrated the important contribution of local connections for the wave propagation. The lateral inhibition is necessary for the localization of the active site on the auditory cortex. Furthermore, for the bifurcation of synchronization - desynchronization of the wave propagation depends on the balance between the neighboring excitatory connections and the lateral inhibition.
These results indicate sounds were represented by dynamic spatio-temporal neural activities in the auditory cortex and propose a novel way of feature and articulation in the auditory recognition. Less
|
