2001 Fiscal Year Final Research Report Summary
Stochastic Resonance for Coupled Systems and its Application to Signal Processing
| Project/Area Number |
12650064
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
Engineering fundamentals
|
|---|
| Research Institution | KYOTO UNIVERSITY |
|---|
Principal Investigator |
IGARASHI Akito School of Informatics, Kyoto University, Associate Professor, 情報学研究科, 助教授 (00115784)
|
|---|
| Project Period (FY) |
2000 – 2001
|
| Keywords | stochastic process / stochastic resonance / noise / signal processing / neural network / retrieval / storage capacity |
|---|
| Research Abstract |
Stochastic resonance (SR) phenomena in coupled systems are investigated with the use of computer simulations and applied to signal processing.
First, we pay attention to the linearly coupled double-well systems, and simulate them with computers. It is observed that for certain values of the coupling constant and noise intensity, signal to noise ratio (SNR) has a maximum, and the maximum value of SNR is greater than that of the uncoupled system.
Secondary, ROC (Receiver Operating Characteristics), which characterized the efficiency of the system in signal processing is obtained. The results show that the ROC for the coupled systems is improved better than that for the uncoupled systems and that SR can be applied to signal processing.
Thirdly, FitzHugh-Nagumo (HN) model, a simplified version of Hodgkin-Huxley model, which is obtained from the analytic fit of voltage of membrane observed by biological experiments, is chosen. We investigate the globally coupled HN model with computer simulation, paying special attention to SR and associative memories with the use of coupling determined by Hebb rule. We show that the probability of the retrieval of a pattern has a maximum for a certain value of noise intensity.
Finally, we use a model for neuron, which is described with a map of a continuous variable for the internal state of the neuron, which is easily treated than HN model governed by differential equations and is less simple than the Hopfield model, where the neuron takes only two kinds of states. We investigate SR phenomena of this system and associative memories with the use of the couplings determined by Hebb rule. The storage capacity of the system is also investigated.
|
Research Products
(4 results)
