基于降噪源分离的脑电信号消噪方法

Denoising method of EEG signal based on tangent function of denoising source separation

为了解决采集的脑电信号中常含有工频、心电、肌电和眼电等多源干扰问题,提出一种基于降噪源分离的脑电信号消噪方法.首先,该方法经过小波分解重构,消除高斯噪声完成预处理;然后,根据脑电信号的非高斯性,用正切函数进行降噪源分离,将含干扰的脑电信号逐次迭代提取得到分离信号作为消噪结果;最后,引入相关系数检测消噪效果.实验结果表明:经过降噪源分离提取得到的分离信号之间呈现弱相关性,而目标分离信号与源信号具有强相关性,可有效去除脑电信号中的心电和眼电伪迹.

The captured electroencephalogram(EEG) signals usually contain multiple sources of interference,such as power frequency,electrocardiogram(EKG),electromyogram(EMG),vertical electrooculogram(VEOG) and so on.In this paper,a denoising method of EEG signals based on tangent function noise source separation was proposed.Firstly,the EEG signals were decomposed by wavelet transform to eliminate the Gauss noise for pretreatment.Secondly,according to the non-Gaussianity of EEG signals,the tangent function was introduced to separate the noise source,and the interfering EEG signals were iteratively extracted generation by generation to obtain the separated signals as the denoising result.Finally,the correlation coefficient was proposed to verify the denoising effect.The experimental result reveals that a weak correlation exists between the isolated signals obtained by noise separation,while the target separated signals show a strong correlation with the source signals,which means that EKG and VEOG artifacts from EEG signals are removed effectively.