一种改进的奇异值分解脑电去噪方法

An improved SVD method for EEG denoising

运动想象是目前脑机接口研究中的一种重要的手段,然而混有大量的生理噪声的脑电信号对于脑机接口的研究有着很大的影响。文章在传统的奇异值分解(Singular Value Decomposition)去噪脑电信号的方法上提出一种改进的奇异值分解方法。针对所采集到的脑电信号,通过敏感因子选取脑电信号的敏感分量,定位因子定位相应的奇异值以实现对脑电的时频信号的重构,以此来去除信号中的噪声,提取有效的脑电信号。通过对握力想象脑电数据的分析结果表明,相比传统的奇异值分解(Singular Value Decomposition)方法,文中所采用的方法能够更加有效的去除脑电信号中的噪声信号,证实了该方法的有效性。

Motor imagery is an important method in brain computer interface research,but EEG signals with large amounts of physiological noise have great influence on brain computer interface research. In this paper,an improved singular value decomposition（ SVD） method is proposed for the traditional singular value decomposition（ SVD） denoising of EEG signals. According to the collected EEG,EEG signal sensitive component selection by sensitive factor,location factor corresponding location of singular value to realize the reconstruction of time-frequency signal of EEG,in order to remove the noise in the signal,it extracts the signal effectively. Through the analysis of imagery EEG data on handgrip strength results show that compared to the conventional SVD method,this method can remove the noise signal of EEG signals more efficiently,confirming the validity of the method.