Grant-in-Aid for Scientific Research (C)
|Allocation Type||Single-year Grants|
|Research Institution||Kyushu Institute of Technology|
INOUE Katsuhiro Kyushu Institute of Technology, Faculty of computer Science and Systems Engineering, Department of Control Engineering and Science, Associate Professor, 情報工学部, 助教授 (00150516)
MAEDA Makoto Kyushu Institute of Technology, Faculty of computer Science and Systems Engineer, 情報工学部, 助手 (00274556)
KUMAMARU Kousuke Kyushu Institute of Technology, Faculty of computer Science and Systems Engineer, 情報工学部, 教授 (30037949)
今井 純 九州工業大学, 情報工学部, 助手 (50243986)
|Project Period (FY)
1994 – 1996
Completed(Fiscal Year 1996)
|Budget Amount *help
¥2,100,000 (Direct Cost : ¥2,100,000)
Fiscal Year 1996 : ¥400,000 (Direct Cost : ¥400,000)
Fiscal Year 1995 : ¥400,000 (Direct Cost : ¥400,000)
Fiscal Year 1994 : ¥1,300,000 (Direct Cost : ¥1,300,000)
|Keywords||Electroencephalogram / EEG Sleep Stages / Human Activity Level / Event Related Potential / Evoked Response / Identification / Signal Processing / Stochastic System / 聴覚誘発電位 / 睡眠ステージ / ニューラルネット|
In this research project, we have performed studies on the development of a discrimination and identification method of human activity level (sleep level and arousal level) based on information processing of electroencephalogram (EEG) signals and obtained following research results.
1.Automatic determination system of human EEG sleep stages
We have developed an automatic stage determination system with graphical user interface based on the wave-shape recognition method. By this system, we obtained about 80% accuracy compared with clinician's scoring.
2.Feature extraction of sleep stages based on EEG generating models
The EEG generating mechanism was modeled by a damped system excited by impulse input process which subjects to Poisson process whose amplitude has a transition property. Furthermore, a Maximum A Posteriori (MAP) method was modified to be applicable to the model. We could then extract useful information directly related to sleep stages by using our method.
ng of EEG based on artificial neural network (ANN)
ANN was applied to EEG signals as the predictor for alpha-waves, high voltage delta-waves and low voltage fast waves. Classification of unknown waves among the above mentioned class were possible by evaluating the infinity norm between ANN output and original signals.
4.Investigation of event related potential (ERP) under visual cognitive task
The measurement system was developed using 2 computers. One is used for visual stimulation, the other is used for data acquisition. Investigating EEG signals under the oddball task (2 kinds of color or 2 kinds of character pattern), it was confirmed that wave patterns were changed by stimulation only at the low arousal level. We also tried to classify the single trial ERPs by using statistical pattern recognition method. This was not easy, because P300 related to vanishing time of stimulation exists in all data (target and non-target stimulus).
5.Quantitative evaluation method of stationarity of EEG signals as the features of human activity level
We have explored the possibilities of the quantitative evaluation based on wavelet transform. It has been confirmed that EEG waves whose duration is 0.5 [sec] could be classified by using wavelet transform and statistical pattern recognition method. These results suggest the possibilities of segmentation for stationary period. Therefore, we will further investigate to find out quantitative evaluation methods related to the segmentations. Less