摘要
由于癫痫的发作一般呈突发性，因在目前还做不到对癫痫发作的有效预测。本文利用高阶统计量对２０例癫痫发作前的两段脑电信号分别建立ＡＲ模型并提取特征，然后通过对这些信号作模式识别，发现在癫痫发作前６ｓ内，脑电信号中已蕴含有在时域看不到的异常变化，从而为癫痫的预测提供了可能。本文对这些脑电信号进一步做双谱分析。
In order to study possibility of epilepsy prediction, we used high order statistics to set the AR model for EEG segments shortly before epileptic explosion, Then we used the coefficients of AR model to complete the pattern recognition. The results showed that abnormal changes were noticed shortly before epileptic explosion, but such changes were invisible in EEG waveform. Bispectrum analysis of these EEG signals showed similar result of pattern recognition. This method was found useful for epilepsy prediction.
出处
《中国生物医学工程学报》
CAS
CSCD
北大核心
1998年第2期111-117,共7页
Chinese Journal of Biomedical Engineering
关键词
癫痫
高阶统计量
AR模型
模式识别
脑电图
Epileptic explosion
Hihgh order statistics
AR model
Bispectrum
Pattern recognition
