基于最大信噪比盲源分离的脑电信号伪迹滤波算法

摘要

Electrocardiography(EKG) and electro-oculogram (EOG) are the most common interference in electroencepha-logram(EEG), and this paper presents an algorithm to filter artifacts in EEG based on blind source separation of maximum signal noise ratio. The algorithm takes separation matrix as a variable to establish the objective function of original signal to noise ratio, looking for the separation matrix which makes the objective function reach a maximum (or minimum ) value.Then the estimated signal can be obtained through the separation matrix.The algorithm has the following steps in the implementation process. Firstly, part noise in the original EEG is removed through wavelet transform. Then the denoised EEG which still contains EKG and EOG is separated by the algorithm to filter artifacts based on blind source separation of maximum signal noise ratio. And correlation coefficient is introduced to verify the consistency between the outputs of blind source separation and the original signals. The experimental results indicate that the cross-correlation coefficient between the output signals of blind separation dropped significantly compared with the original signals.Thus the algorithm is effective for removing EKG and EOG in EEG.%心电和眼电伪迹是脑电信号中最常见的干扰,本文提出一种基于最大信噪比盲源分离的伪迹滤波算法.该算法以分离矩阵为变元建立源信号的信噪比目标函数,寻找能使目标函数达到极大(或极小)值的分离矩阵,进而通过分离矩阵求得估计信号.算法的实施过程是,首先利用小波变换去除在原始脑电信号中的部分噪声,然后用基于最大信噪比盲源分离的伪迹滤波算法对消澡后仍含有心电和眼电的脑电信号进行盲信号分离,并引入相关系数验证盲信号分离输出与源信号的一致性.实验结果表明盲分离后各输出信号间的互相关系数较分离前大幅下降,从而证实了算法对于心电和眼电伪迹分离的有效性.