基于扩展Informax算法的脑电信号伪差分离研究

提出了一种基于扩展Informax算法的脑电伪差检测和分离方法，首先应用扩展Informax算法将脑电信号分解为相互独立的分量，并计算其解混矩阵和混合矩阵，通过分析分解结果发现；脑电信号中的伪差成分总是作为一到两个独立分量被分解出来，并且这些分量的空间分布与特定伪差类型相对应，因此，根据混合矩阵提取独立分量的空间分布特征，即可检测出伪差成分，进而重构出无伪差的脑电信号，分离眨眼伪差的实验结果也充分证明了这一方法的有效性。

Study on Removing Electroencephalographic Artifacts Based on Extended Informax Algorithm

Based on the extended informax algorithm, an electroencephalographic (EEG) artifacts removing method is proposed. First, the EEG signal is decomposed into independent components by the extended informax algorithm, and then the mixing and unmixing matrix is calculated. From the decomposing results, it is found that the artifacts always are involved in one or two independent components, and their distributions relate to the type of artifact. From the mixing matrix the distribution of each independent component can be extracted, then the artifact components can be detected. After the artifacts are removed, the clean EEG signal can be reconstructed. Applying to blink artifacts removing, the results show that the proposed method is quite promising.