To extraction the feature of EEG signals for mental task recognition

Electroencephalogram (EEG) signals are widely used in brain-computer interface (BCI) recently. By classifying the signals, mental tasks in the brain are available to be estimated and then the results can be used in the communication between human and external devices or robots. Many classifiers, such as support vector machine (SVM), multi-layer perceptron (MLP), and self-organizing map (SOM), etc., have been proposed, however, the crucial role in these techniques is how to extract the features of the EEG signals. In this paper, an efficiency feature extraction method is proposed. It includes a series processes such as windowing, Fourier transformation, frequency band filtering, and nonlinear normalization. Additionally, several supervised classifiers such as MLP, SVMs, and “class proximity self-organizing map (CP-SOM)” proposed by Saito & Harton are adopted to verify the effectiveness of proposed method. Using a benchmark data and the original data measured by the EEG sensor EPOC, mental task recognition experiments were executed and the priority of the proposed method for all classifiers was confirmed.