文章摘要

文章信息/Info

Author(s):: XU Yu; ZHENG Wei; CHENG Yi; （Jiangsu University of Science and Technology, Zhenjiang 212000, China）

Keywords:: Signal processing and analysis; Denoising; Mode decomposition; Root mean square error; Correlation coefficient; Robustness

摘要:: 为消除脑电信号中眼电伪迹对信号分析带来的干扰，本研究提出一种核独立成分分析（kernel independent component analysis，KICA）与自适应噪声完备经验模态分解（complete ensemble empirical mode decomposition with adaptire noive，CEEMDAN）相结合的方法去除眼电伪迹。首先，使用CEEMDAN方法对脑电信号进行分解;再对分解后得到的模态分量进行KICA降维;最后对核独立分量进行样本熵计算，利用阈值对眼电分量进行判别，去除眼电伪迹，重构出不含眼电伪迹的脑电信号。通过多个公共数据库来对本研究方法进行验证，并与经典ICA方法进行对比，实验结果表明，本方法对于脑电信号中的眼电伪迹去除具有较好的实际效果。

Abstract:: In order to eliminate the interference brought by the eye-electric artifacts in EEG signals to the subsequent signal analysis, we proposed a combined kernel independent component analysis (KICA) and complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) method to remove the eye-electric artifacts. Firstly, the EEG signal was decomposed by adaptive white noise complete empirical modal decomposition; then the decomposed modal components were analyzed by KICA; finally, the sample entropy was calculated for the kernel independent components, the threshold was used to discriminate the EEG components to remove ocular artifacts reconstructed the EEG signal without EEG artifacts. The proposed method was validated by multiple public databases and compared with the classical ICA method. The experimental results show that the proposed method has good practical effects on the removal of EEG artifacts from EEG signals.