Research On EEG Signal Analysis Techniques Based On Blind Source Separation

EEG (Electroencephalography) contains a wealth of information, which is oftenapplied in brain computer interface in engineering and disease diagnosis in clinic. Theability to analyze the EEG signal and extract information of EEG features effectively andquickly is an important criterion to judge whether the performance in brain computerinterface and disease diagnosis is good or not. In this paper, after analyzing the currentdomestic and international methods about EEG signal processing, one method is used toanalyze EEG signal applying blind source separation algorrithm which is widely used inthe field of signal processing and with P300and motor imagery EEG signals as the objectof the research respectively.P300’s intensity is so weak that it is easy to suffer from interferences of environmentand artifacts of blink, ECG, and EMG and normally submerged in gathered EEG data. Inorder to separate P300from interferences speedily and efficiently, this paper analyzes thefeature of P300in time, frequency and spatial domain. A method is proposed based onBSS (Blind Source Separation) to extract P300, combining coherence average, wavelettransforming and BSS. A new technique is described to automatically select independentcomponent corresponding to P300from multiple source signal estimation component afterblind source separation. The performance of three BSS algorithms, that are Informax,FastICA and AMUSE, are compared in the process of P300extraction. Experimentalresults verify that the proposed method based on BSS has an obvious improvement inP300extraction compared with the other methods using temporal and frequencycharacteristics of P300only.In order to extract feature of the motor imagery EEG signal accurately and efficiently,this paper analyzes the feature of the motor imagery EEG in time, frequency and spatialdomain firstly. Then an approach is put forward that extract the motor imagery EEG fromtime, frequency and spatial domain, with preprocessing of wavelet transforming and usingspatial filter obtained from SOBI and ITFE. And take the energy of motor imagery EEGextracted as it’s feature. Experimental results verify that the proposed method in this paper to extract feature of the motor imagery EEG has certain superiorities, and the spatial filterobtained from SOBI and ITFE can reflect more real brain activities.