Featuer Extraction And Classidication Of Attention Related EEG

Electroencephalogram(EEG) is the overall reflection of electrophysiologicalactivities of the nerve cells in the cerebral cortex or scalp surface, and it contains alarge amount of information of physiological and disease. The extraction andprocessing of EEG could provide clinical diagnosis basis for physicians, which alsoprovide effective treatments to some brain diseases. Attention deficit disorder(ADD) isa common disease in the overlap area of the physical and mental diseases, whichoccurs mostly in the6~16years youth. The study of ADD treatment system based onEEG feedback is popular in recent years. How to classify attention relatedElectroencephalographic quickly and effectively is the key issue of this researchFirstly, attention related EEG signal acquisition experiments are designed, and theexperimental data is collected from fourteen subjects in a relatively quiet laboratory.Secondly, EEG signals are preprocessed and decomposed by wavelet transforms (WT),empirical mode decomposition (EMD) and local mean decomposition (LMD). Thenprogramming to realize the above algorithm in Matlab, and optimize some parameters.In the WT research, the EEG signals are decomposed to eight layers by using db4wavelet, the energy of the4rhythms(δ、θ、α、β) are extracted as the feature. The energyof the first four IMF components after EMD, and the energy of the first four PFcomponents after LMD are extracted as the feature. Finally, combined with supportvector machine (SVM) theory, the above three feature matrixes are used as input ofSVM whose parameters optimized, the output of SVM are the classification results ofdifferent signals.The results show that, the three methods above all achieved a good effect inattention related EEG signals classification, the classification accuracy of WT is91.43%, that of EMD is89.29%, and which is90%with LMD.