| Project/Area Number |
17605009
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
睡眠学
|
|---|
| Research Institution | Kyushu Institute of Technology |
|---|
Principal Investigator |
INOUE Katsuhiro Kyushu Institute of Technology, Faculty of Computer Science and Systems Engineering, Professor, 情報工学部, 教授 (00150516)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
MAEDA Makoto Kyushu Institute of Technology, Faculty of Computer Science and Systems Engineering, Research Associate, 情報工学部, 助手 (00274556)
KUMAMARU Kousuke Kyushu Institute of Technology, Faculty of Computer Science and Systems Engineering, Professor, 情報工学部, 教授 (30037949)
|
|---|
| Project Period (FY) |
2005 – 2006
|
| Project Status |
Completed (Fiscal Year 2006)
|
| Budget Amount *help |
¥3,600,000 (Direct Cost: ¥3,600,000)
Fiscal Year 2006: ¥1,100,000 (Direct Cost: ¥1,100,000)
Fiscal Year 2005: ¥2,500,000 (Direct Cost: ¥2,500,000)
|
|---|
| Keywords | Biosignal Processing / Sleep / Autonomic Rhythm / Independent Component Analysis / Wavelet Analysis / Peak Frequency Analysis / Quasi-AR Model / Feature Extraction / 独立成分解析 |
|---|
| Research Abstract |
In this research, feature extracting methods related to the evaluation of the quality of the human sleep were investigated, and the following results were obtained.
1. Feature extraction method based on wavelet analysis
The adjusting rule of the damping coefficient of Gabor function in the modified wavelet transform was modified to the rule based on nonlinear function (sigmoidal function). As a result of processing the data sampled with sampling frequency 500Hz, it was confirmed that a peak frequency time series were able to be estimated more stably. Moreover, it was confirmed that distribution of the power in the B wave band strongly related to the sleep EEG stages.
2. Independent component analysis of multi channel EEG signals
The independent component analysis (blind source separation method) was applied to the analysis of the multi channel EEG wave. It was confirmed that the number of source signals could be estimated from the separation signals and that the outline position of the sou
… More
rce signals could be estimated based on the correlation between the separation signals and original EEG signals. Usually a few EEG signals (C3, C4, etc.,) are measured at the sleep research. The influence of environment that the number of sensors is fewer than the number of source signals is under investigating.
3. Autonomic rhythm extraction method (ECG analysis)
The R-R interval time series was extracted from the ECG signal, and the HF/LF component was extracted based on the AR spectral analysis every one minute, and the correlation with the sleep EEG stage was investigated. As a result, it was confirmed that the instability of the rhythm extraction was caused by the phenomenon that the heart beats fluctuated by the twice or more even in the period of 20 seconds by the influence of the body motion etc.
4. Other feature extraction methods
Feature extraction method based on multidimensional directed information, Quasi-AR model and generalized normal distribution were investigated. These methods were confirmed to be effective for the evaluation of the quality of sleep, because the information flow and positions of source in the brain and the short period (about 15 minutes) rhythm were able to be extracted. Less
|
