基于PCA、ICA的脑电伪迹消除研究

【作者】 陈立伟；

【导师】 尧德中；

【作者基本信息】 电子科技大学 ， 生物医学工程， 2004， 硕士

【摘要】 脑电(EEG)中蕴涵着丰富的生理、心理及病理信号,脑电信号的分析和处理无论在临床对一些脑疾病的诊断和治疗,还是在生命科学的研究领域都是十分重要的。许多测量到的生理信号往往是若干独立成份的线性加权迭加得到的,如诱发脑电总是被自发脑电所淹没,而常常伴有心电、眼电(EOG)、头皮肌电和工频电源等的干扰。从观测信号中分离出具有真实生理意义的各自独立的成份和实现伪迹的消除具有重要意义。在现行的信号分析方法中,主成份分析(Principal Component Analysis-PCA)是基于二阶统计量把信号分解成若干相互正交的信号,独立成份分析(Independent Component Analysis-ICA)作为基于高阶统计量的信号处理方法,能够抑制高斯白色噪声和有色噪声,并能分解出相互独立的非高斯信号,在信号处理界引起广泛关注,该技术也被迅速应用到生物信号处理领域。本论文主要分析了ICA理论及其算法,提出了将其应用于眼电伪迹的分离和消除,得到了比较理想的结果,与此同时也介绍了采用PCA方法对脑电中眼电伪迹消除的处理过程和结果。本论文具体完成的工作:1、利用PCA、 ICA 对EEG中的EOG的自动消除,实现了波形还原时平滑连接,不出现锯齿波动。2、NEUROSCAN 的CNT、 AVG 等格式数据的读写,及其与DAT等格式之间的转换,为脑电处理系统软件的研究奠定了一定的基础。更多还原

【Abstract】 The analysis and processing of EEG signal are very important, not only in clinic diagnosis and treatment of some brain diseases, but also in the life science research field.As ERP is contaminated with EEG, Various recorded biomedical signals practically are mixture of different independent source signals , artifacts and noises, such as EOG、ECG、EMG and other noises. It is very important for us to extract the meaningful signals from such a mixture.In traditional method of signal analysis, Principal Component Analysis-PCA decompose recordings implements into mutual orthogonal signals. It is based on two order cumulate. Independent Component Analysis-ICA is a high order cumulate signal analysis method, and it can suppress Gauss noise and Colored noise, and can separate independent None Gauss signal . ICA has an important value in the biomedical signal processing and is worthy of being completely studied.In this paper, we analyse ICA theory and algorithm, and use this method to remove EOG artifacts from EEG recordings. The experiment results show that it is a promising method. We also study PCA method, and found its computational efficiency in removing EOG artifacts from EEG.The innovated works we have finished are as following:Present the methods using PCA and ICA to remove the EOG artifacts in EEG signals automatically, and resume the original signal without the saw tooth fluctuation. 2、The software can fulfill the function of reading and writing the data of NEUR SCAN’ files such as CNT and AVG, and translating between the CNT、AVG and DAT. These are the basis of further research in EEG signal processing system software.更多还原