Classification of EEG signals by ICA and OVR-CSP

Signal processing of Electroencephalography (EEG) plays an important role in brain-computer-interface (BCI) system. Therefore, it is a key in selection of suitable methods. This paper proposed a new method by combing Independent Component Analysis (ICA) with One Versus the Rest Common Spatial Patterns (OVR-CSP) to improve the classification performance. Firstly, EEG signals were filtered with FIR bandpass filter 8-30HZ. Secondly, relative frequency band signals (i.e. μ and β rhythm) were decomposed into independent components to obtain the solution matrix by ICA, and the EEG signals were reconstructed by the main components for improving the signal-to-noise ratio. In order to capture the essential structure of the data, OVR-CSP was used to extract the feature of EEG signals and reduce data dimensions. Finally, using Support Vector Machines (SVM) to classify the feature. The experiment result shows that the average accuracy rate of EEG signals based on motor imagery by the proposed method could achieve 95.555%. It can be understood that ICA is a very effective method to remove artifacts in EEG.