Study On The Removal Of The Artifacts From EEG Data Based On The ICA And WCCA

Electroencephalogram (EEG) is a kind of electrical activities recorded by the electrodes placed on the scalp surface. The amplitude of EEG is very small, with intensity around microvolt. Artifacts such as electrocardiogram(ECG), electromyography (EMG) and electro-oculogram (EOG) often contaminate the EEG signal. Among these contaminating sources, EOG is the major one. They distort the electric field distribution of actual brain activities over the scalp, especially surrounding the eyes. It is very difficult to read and analyse the EEG signal in these contaminations especially for automatic analysis and diagnoses. How to reduce or reject these contaminations from EEG has become a very important issue. Almost all users must take into account the effect of ocular activities on EEG, and the method of EOG removal should be reported in publications. With the availability of digital EEG, it has become a desirable procedure to correct artifacts automatically by computer.In this article, the following EOG removal methods, which are based on scalp EEG data are presented and discussed.1)First, the article introduced the characteristics of common artifacts in EEG, analysed the widely used methods of EOG removal, especially the Independent Component Analysis(ICA), Wavelet Transform (WT).2) The article realized the algorithm FastICA, and it is used to reject the EOG artifact in the EEG Although ICA method can reject the EOG artifact in EEG, it needs manual intervention and can’t achieve automatically removing of the EOG artifact.3) A new method based on the Canonical Correlation Analysis(CCA) and WT is presented, since the ICA method cannot reject EOG artifact automatically. We can determine the scale of wavelet decomposition by using CCA method, accordingly realizing the automatically removing of EOG artifact in EEG What’s more, the article has certified the WCCA method by the analysis of time-domain characteristics and spectral characteristics of processed EEG The article evaluate the de-noising result from SNR and RMSE.