【作者】 白树林；

【导师】 谢松云；

【作者基本信息】 西北工业大学 ， 电路与系统， 2006， 硕士

【摘要】 脑电图(electroencephalogram,EEG)是脑神经细胞的电生理活动在大脑皮层或头皮表面表现出的电现象。一般来说,脑电变化可分为两类:即诱发电位响应和自发电活动。研究表明,脑电信号具有背景噪声强信号幅度微弱、非平稳性和随机性强、频域特征比较突出等特点。因此,脑电信号的分析与处理仍然是一项非常具有挑战性的课题。 灰色系统理论(Grey System Theory)是由我国学者邓聚龙教授于1982年在国际上首先提出的。对于非典型规律的信号(如非平稳、非高斯分布、非白噪声),灰色方法与其它的一些按统计规律和先验规律来处理数据的方法相比,具有明显的优势。 本文在初步分析了脑电信号处理方法以及灰色系统理论的基础上,考虑到脑电信号的非平稳性和随机性强、频域特征比较突出,结合灰建模对于建模数据无特殊性要求等特点,提出了脑电信号分析与处理的一种新方法——将灰建模理论应用于自发脑电特征的提取中。同时对于采用灰建模方法提取的脑电特征,采用机器学习理论中基于实例的k-近邻算法对实测脑电信号进行了分类识别。研究结果表明,在脑电信号处理中使用灰建模方法提取脑电特征并采用k-近邻算法进行分类决策,在理论上是可行的、有效的。同时该方法也为进一步的脑功能模式识别研究提供了良好的理论基础。本文主要完成了以下工作: (1)脑电信号的灰色GM(1,1)建模; (2)模型参数估计及脑电特征提取: (3)分析比较两种状态,即睁眼和闭眼时脑电特征参数a,b的不同,并给出比较结果; (4)采用基于实例的k-近邻算法对未知脑电信号(睁眼和闭眼)模式进行分类决策,并给出在选取不同参数时算法的分类性能以及改进方法。 本文在MATLAB环境下实现了所有的分析处理工作。研究结果表明,本文提出的方法是可行有效的。更多还原

【Abstract】 Electroencephalogram (EEG) is electrical phenomenon represented from pallium or scalp surface of the electrical activities about brain nerve cell. In general, EEG could be divided into two types, evoked potentials and spontaneous EEG. Many researches show that EEG is characteristic of stronger background noise and fainter signals amplitude, higher non-steady and randomicity, relatively prominence of frequency character, et al. Therefore, the analysis and processing of the EEG is still a very challenge subject.Grey System Theory (GST) is created by Chinese scholar professor Deng Julong in 1982. For non classic regularity signals such as non-steady, non-Gaussian and non-white noise, GST has the obvious advantages than other processing methods according to statistical and transcendent regularities.In this paper, in the basis of analyzing the processing methods of the EEG and GST, meanwhile, considering that higher non-steady and randomicity, obvious frequency character, and that for the modeling data grey modeling has not specific demands, a novel method is put forward for the analysis and processing of the EEG signals, i.e. the theory of grey modeling is used into the feature extraction of the Spontaneous EEG At the same time, using the EEG features from grey modeling approach, the classification of the recognized EEG signals are performed by the k -Nearest Neighbor (k-NN) Algorithm from instance-based learning method (IBL). The results of the research show that in the processing of the EEG signals, it is applicable and available by using grey modeling method to perform feature extraction and making classification according to k-NN Algorithm. Meanwhile, this method also provides well theory foundation for further research of the pattern recognition of the EEG. Specifically, in this paper, the following work is performed:(1) Modeling GM (1, 1) for EEG signals.(2) Parameters estimation of the model and EEG feature extraction.(3) Feature parameters a, b for two states (eyes-open and eyes-closed) are analyzed and compared and the comparing results are given.(4) The classification of unknown EEG patterns (eyes-open and eyes-closed) is performed by using k -NN algorithm from instance-based learning methods and the classification performance of the algorithm under the different parameters and improving methods are presented.The whole analysis work is completed under the MATLAB environment. The eventual results of the research show that the method proposed in the paper is applicable and available.更多还原