The Analysis And Classification Of EEG Bispectrum Slice

EEG is a brain bioelectrical activity，which can reflect the cerebral cortex, extractand analyse EEG, and it can also provide a good medical clinical diagnosis support, aswell as a good effect on the treatment of brain diseases. On the other hand, EEG is avery complex non-Gaussian, nonlinear stochastic process, using the time domain orfrequency domain analysis can not accurately characterize the signal characteristics,and it is not conducive to extract the transient waveform EEG feature, but thehigh-order spectrum analysis methods for the use of EEG signal processing researchdoes have a good effect.In this paper, we invited a dozen of volunteers. Each volunteer was told to makethree different tests of attention EEG experiments, then the experiment data werecollected and stored, Then the various experimental data were filtered, down-sampled.and Staged. By using the pre-processing of the EEG data, each individual experimentalthree-dimensional spectra was obtained, as well as the EEG bispectrum slice of everyexperimental data. Then we extract the eigenvalues of EEG bispectrum slice, and useSVM, KNN, PNN classification methods to process the spectral slice eigenvalues toobtain the EEG classification accuracy on different experimental conditions. The resultshows that the SVM classifier has the highest classification accuracy. According to thedifferent OQ experimental bispectrum characteristics, comparisons among C3electrode and other13electrodes were made, the result shows that the signals ofdifferent slice of brain electrode have different bispectral characteristics. Finally, thepower spectral characteristics method is used to compare with the value method toimprove the classification results between focus attention and distraction of EEG.