Feature Extraction of Brain-Computer Interface based on Second Order Blind Identification and ICA

Abstract

Blind Source Separation (BSS) is a technique for recovering unobservable source signals from their mixtures such as EEG signals. Second Order Blind Identification (SOBI) is one of algorithms of BSS and it is a suitable method for analyzing EEG components for its typical non-stationary and weakness. Independent component analysis (ICA) is used to select non- Gaussianity component of mixed signals. In this woke we study feature extraction of EEG signals for classifying spontaneous mental activities in brain computer interface (BCI). Three subjects participated in the BCI experiment which contains two mental tasks including imagination of left hand and right hand movement. After preprocessing, SOBI and ICA were applied to achieve special components of EEG signals. Then, the features were extracted and Linear Discriminant Analysis (LDA) was used to classify the feature extracted. The result shows that SOBI plus ICA is an effective feature extraction method for subjects whose accuracy was lower when using tradition feature extracted methods.