Brain Computer Interface Design Based On Mental Task EEG

Brain-Computer Interface (BCI) using Electroencephalogram (EEG) signals is a hot research subject in recent years. It must be mentioned that BCI based on EEG is not to explain EEG, but to identify EEG. The structure of a BCI can singly be divided into two parts: the front is EEG signal's character extraction and classification; the other is combination and application with necessary hardware. Here the focus is the front, designing a high classification BCI only using simple sensibilities mental task EEG signal and indication a method using this BCI to sort many kinds of tasks.First, ICA was used to remove the noise in EEG data. Second, extracting EEG data characters of three different metal tasks used AR model coefficient, the approximate-entropy and the wavelet-entropy. Among them, the wavelet-entropy not only showed the signal's complex in time domain, but also showed the signal's other character in frequency domain. So it suit to deal with EEG signal. Finally, Classifier adopt BP neural network, RBF neural network and support vector machine (SVM). SVM achieve the highest accuracy comparing these three classifier's result.The experimental result shows that the feature extraction method used wavelet-entropy and SVM can get a high accuracy classification (about 95%). The result indicates that...