基于EEMD-ICA的脑电去噪算法研究

脑电信号(EEG)是脑神经细胞的电生理活动在大脑皮层的反映,但采集到的脑电信号一般都含有大量噪声。为了保留有效信息同时消除尽可能多的噪声,提出通过构造虚拟通道将集合经验模态分解与独立成分分析相结合的脑电信号去噪方法。首先,对脑电信号进行EEMD分解得到固有模态函数(IMF)分量,根据相关性准则筛选含噪声成分多的IMF分量构造虚拟通道进行ICA去噪;然后,将消噪后的结果与含信号成分多的IMF分量进行重构再次ICA去噪,得到最终去噪信号。为了验证EEMD-ICA去噪方法的有效性,以信噪比、均方根误差作为评价指标,将该方法与小波去噪法、EEMD去噪法、ICA去噪法进行比较,结果表明,EEMD-ICA去噪后的信噪比高于其它方法,均方根误差小于其它方法,综合分析该方法能更好地消除噪声。

Denoising Method of EEG Signal Based on EEMD-ICA

The electroencephalograph( EEG) signal is the electro physiological activity of brain cells,it is reflected in the scalp surface. However it is usually interfered by noises during signal acquisition process. In order to reserve the effective information and eliminate as much noise as possible,a method of ensemble empirical mode decomposition( EEMD) combined with independent component analysis( ICA) is introduced. Firstly,the EEMD decomposition can get a certain number of intrinsic mode function( IMF) of EEG signals. The virtual channel is reconstructed by the IMF components with more noise components which are selected based on correlation coefficient and de-noised by ICA algorithm. Secondly,the denoised results and the IMF components with multiple signal are reconstructed. Finally,the reconstructed signal is denoised by ICA again,and the final denoised signal is obtained. The experimental results show that the mentioned method has better signal-to-noise ratio and smaller RMSE than the other denoising methods,including wavelet denoising,EEMD denoising and ICA denoising. It shows that the mentioned method can denoise better.