| Project/Area Number |
13555113
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (B)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 展開研究 |
|---|
| Research Field |
System engineering
|
|---|
| Research Institution | Kobe University |
|---|
Principal Investigator |
KOTANI Manabu Kobe University, Faculty of Engineering, Associate Professor, 工学部, 助教授 (30215272)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
OZAWA Seiichi Kobe University, Graduate School of Science and Technology, Associate Professor, 自然科学研究科, 助教授 (70214129)
OGAWA Kazuhiko Kobe University, Faculty of Engineering, Associate Professor, 工学部, 助教授 (30252802)
|
|---|
| Project Period (FY) |
2001 – 2003
|
| Project Status |
Completed (Fiscal Year 2003)
|
| Budget Amount *help |
¥5,700,000 (Direct Cost: ¥5,700,000)
Fiscal Year 2003: ¥1,100,000 (Direct Cost: ¥1,100,000)
Fiscal Year 2002: ¥2,400,000 (Direct Cost: ¥2,400,000)
Fiscal Year 2001: ¥2,200,000 (Direct Cost: ¥2,200,000)
|
|---|
| Keywords | Acoustic diagnosis / Leakage sound / Independent component analysis / Neural network / Kernel method / Feature extraction / Evolutionary computation / 音響 / 設備診断 / ガス漏洩 / モジュール |
|---|
| Research Abstract |
We have developed the acoustic diagnosis system which is capable of adapting to the dynamic environment.
The major results of the project as follows :
1. Development diagnostic networks
We proposed a novel model of modular neural networks which has the ability to adapt its structure according to the environment. Experiments were performed for an artificial gas leakage device with various experimental conditions to imitate the change of environment for a long term. The discrimination accuracy with the proposed network was observed to be about 93%. Result shows that the proposed model is effective for detection of the leakage sound for the practical use.
2. Independent component analysis and evolutionary computation as feature extraction
We have examined the feature extraction for acoustic signal using independent component analysis and genetic algorithms to obtain the stable diagnosis. We proposed a novel recognition method using features extracted by ICA. The proposed method consists of some modules for each category and a synthesizer. We evaluate the performance of the proposed method for several recognition tasks including acoustic diagnosis. From these results, we confirmed the effectiveness of the recognition method using independent components for each class. The effectiveness of the proposed method were also confirmed for biological instruments.
|
