【作者】 韦保林；

【导师】 罗晓曙；

【作者基本信息】 广西师范大学 ， 电路与系统， 2002， 硕士

【摘要】 脑电（EEG）信号是临床脑疾病诊断、神经生理学、脑科学等研究的重要途径。脑电（EEG）信号中癫痫特征波的自动检测以及脑电信号的预测在临床检测、脑电监护、癫痫等脑疾病的控制与治疗上均有很重要的意义。 脑电（EEG）信号具有高维的混沌特性，混沌信号的预测是当前混沌理论与应用研究的一个重要方向。但是当前的研究多数是针对低维混沌信号，本文对高维混沌特性的脑电信号的预测研究作初步探讨。人工神经网络在自适应、非线性映射等方面具有优良的性能，在混沌信号预测中具有广泛的应用。而Volterra级数展开式具有高度非线性，大量的非线性系统可以由Volterra级数表征。因此，本文采用径向基函数神经网络和三阶Volterra级数滤波器分别构造脑电信号的两个预测模型；对通常的径向基函数预测网络作了改进，并且采用乘积耦合近似实现三阶Volterra级数滤波器，突破了常规上只能用二阶Volterra级数滤波器的限制。 小波变换是一种时频域分析方法，它具有多分辩率、相对带宽恒定、表征信号局部特征的能力等优点。本文选取适当的小波函数，将信号进行连续小波变换，把脑电信号中的癫痫特征波在不同的尺度下分离出来，然后选取特定尺度下的变换结果，通过阈值判定方法对棘波进行检测。 数值仿真结果表明，本文所研究的两个预测模型以及棘波的小波变换检测方法是很有效的，特别是三阶Volterra级数滤波器预测模型，它对脑电（EEG）信号的预测精度可以达到10^-3数量级。更多还原

【Abstract】 EEG signal is an important method in clinical disease diagnosed,neurophysiology,brain research,etc. Automatic detection of epileptic waves in EEG signal and EEG signal prediction is of great significance for clinical application,long-term monitoring (i.e. EEG Holter) and treatment and control of epilepsy.EEG signal is characterized by high-dimension chaotic. The prediction of chaotic signal is an importance area of chaos theory and its application. However,most existing research results are main about the prediction of the low-dimension chaotic signal,we investigate the prediction of high-dimension chaotic EEG signal in this paper. Artificial neural network has excellent performance in adapting ,nonlinear mapping etc.,and it is widely used to predict chaotic signal. The Volterra series expansion has altitudinal nonlinearly ,most nonlinear system can be expressed by the Volterra series expansion. So we adopt an Radial Basic Function neural network and a three-order Volterra series filter to construct two prediction model which are used to predict the EEG signal. In this paper,we improve the Radial Basic Function neural network and breach the two-order limited by using product-coupling to realize the three-order Volterra series filter.Wavelet transform is an analytical method that unites the time and frequency domain. It has such feathers as multi-resolution,constant relative bandwidth,and the ability to indicate the local features of signal in time and space. After wavelet transform by using proper wavelet basis functions,the epileptic waves can beseparated at different scale,then we can detect the epileptic waves by using the wavelet transform result at the proper threshold value.Simulation results show that:the two prediction model and the method to detached the epileptic waves by using the wavelet transform is effective. Especially,the prediction precision of EEG signal is higher than 10-3 using the three-order Volterra series filter.更多还原