酒精脑电信号降维去噪方法的研究  

Research on Denoising Method of Alcohol EEG Signals

作　　者：赵蕾 白雪梅[1] 胡超 ZHAO Lei;BAI Xue-mei;HU Chao(School of Electronics and Information Engineering,Changchun University of Science and Technology,Changchun 130022)

机构地区：[1]长春理工大学电子信息工程学院

出　　处：《长春理工大学学报（自然科学版）》2019年第6期78-82,共5页Journal of Changchun University of Science and Technology(Natural Science Edition)

基　　金：吉林省自然科学基金项目(20150101013JC)

摘　　要：脑电信号数据维度高,且极易受到噪声干扰。除环境噪声干扰外,自身伪迹干扰对脑电的影响更严重,因此,对采集到的原始数据进行去噪处理是有必要的。常见的脑电去噪方法(如FastICA、奇异性检测、差分谱法等)并不适用于采样点数有限的酒精脑电去噪,经过实验,去噪后的酒精脑电信号特征损失严重。针对酒精脑电数据特点在传统脑电信号预处理方法的基础上,提出了一种基于主成分分析算法(Principal components analysis,PCA)和奇异值迭代分解法(Singular Value Decomposition,SVD)相结合的方法对酒精脑电信号进行去噪。通过PCA算法成功将64维脑电信号降至15维,减少了脑电信号处理的计算量,对提取的主成分进行4次奇异值迭代分解,达到了对脑电信号去噪的同时保留了更多信号特征的目的。EEG data has high dimensionality and is vulnerable to noise interference.In addition to environmental noise interference,self-artifact interference has a more serious impact on EEG.Therefore,it is necessary to denoise the col-lected raw data.Common EEG denoising methods(such as FastICA,singularity detection,differential spectrum meth-od,etc.)are not suitable for ethanol EEG denoising with limited sampling points.After experiments,the characteristic loss of ethanol EEG signal after denoising is serious.Aiming at the characteristics of ethanol EEG data,based on the traditional EEG signal preprocessing method,in this paper,a method combining Principal Component Analysis(PCA)and Singular Value Decomposition(SVD)is proposed to denoise ethanol EEG signal.64-dimensional EEG signal to 15-dimensional is successfully reduced by PCA algorithm;the computational complexity of EEG signal pro-cessing is reduced and the extracted principal components are decomposed by four singular value iterations,which the goal of denoising EEG signal is achieved while retaining more signal characteristics.

关 键 词：酒精脑电信号 PCA SVD 迭代 去噪 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]