Based On Wavelet Decomposition Of The Imagination About The Feature Extration Of Eeg Classification

BCI is a new communication tools between people and computer. The BCI equipment collects the EEG information, extracting the common features of the EEG and classifies it, and then combination of the different thinking activities and the different commands (for example, the mouse up and down movement, etc.), realize the communications between Brain and the external equipment. In order to facilitate those peoples whom they can not care themselves, because they have nerve tissue diseases which lead to the brain signals can not be passed to the muscles, the researches starting the brain-computer interface (BCI) research. In this paper, First pre-processing both the left and right imagine movement C3, C4 electrodes EEG signal , to extract the features of imagine movement EEG signal. In the end, realized automatically classified the brain's electrical signals,causing possibly to become using the EEG to control instrumentation.The main research:(1) Preprocess: short-time Fourier transform converted the time-domain signal to time-frequency signal, using the Spectrogram method in the Matlab toolbox. Deal with the start signals generated by stimulate through line conversion, will get the obvious differences on the lift-right imagine in some frequency ranges.(2) Wavelet analysis: after select the particular function and wavelet decomposition level, deal with the EEG single test data through wavelet decomposition, includes the obvious features of the frequency ranges through the strengthening of stacking repeated features, and then use fisher distance to extraction the feature. Using the stack Repeated means to strengthen the feature of the EEG, which have the obvious characteristics, and using the fisher distance to extract the features.(3) BP neural network training and classification: using the fisher distance to select the eligible sample, input the imaginary EEG and fisher distance as a learning sample to train, and identify the test samples, the identification of k3b data rate to reach 93.1%.