Abstract: In traditional blind source separation algorithms, they usually need two EOG signals as the references to eliminate EOG artifacts. However, when collecting the EOG signals, they will always easily make the subjects uncomfortable, and require manual identification. In order to solve these problems, a FastICA-based method is presented, which can automatically remove ocular artifacts. Firstly, the correlation coefficient between each independent component extracted by FastICA and GFP（Global Field Power） value is calculated. Secondly, compared with these correlation coefficients, the independent component that has the largest absolute value is identified as the independent component of the ocular artifact. Finally, the independent component is set zero to reconstruct the clean EEG signals so that the automatic removal of EOG artifacts is achieved. The 30 cases of experiment EEG data show that this method can quickly and precisely eliminate the ocular artifacts which is completely automatic, preserve the other EEG components, and can be applied in real-time occasions.

Key words: Electroencephalography（EEG）, ocular artifact, Independent Component Analysis（ICA）, automatic removal

摘要： 传统盲源分离算法消除眼电伪迹须用到两个眼电信号作为参考，但在采集眼电信号时易给被试带来不适产生噪声，且识别时需要人为辨别，为了解决这些问题，提出一种基于FastICA的眼电伪迹自动去除方法。该方法先计算出FastICA提取出的各独立成分与GFP（Global Field Power）值的相关系数，再比较相关系数，将其绝对值最大所对应的独立成分识别为眼电伪迹独立成分，最后把该独立成分置零重构干净的脑电信号，实现眼电伪迹的自动去除。通过自采的30例脑电数据实验结果表明：该方法能完全自动地去除眼电伪迹成分并有效保留其他脑电成分，且快速准确，适用于实时场合。

关键词: 脑电信号, 眼电伪迹, 独立成分分析, 自动去除