Study On Extraction Of α Rhythm In Eeg Based On Wavelet Package And Ica

The purpose of EEG signal processing is to extract the hidden or weak patterns from EEG signals in sophisticated noise background. Independent Component Analysis(ICA) can effectively separate the independent source signals from the recorded multi-channel signals, and the separation results will not be affected by spectra overlapping, when some certain assumptions are satisfied. But under the limitation number of EEG signals, the ICA methods can hardly get the good splitting performance.This dissertation firstly introduced the model of ICA, measuring method and optimizing algorithm, and discussed the main judging approaches including Nongaussian-maximization criterion. Besides, the theory and realization of the algorithms of Fast ICA and Infomax are also illustrated in this thesis.Next, the ICA is applied to do the de-noise to the EEG signal, getting rid of the ECG recordings and the power frequency interference among the EEG in the limitation of the number of EEG. In addition, experimental results show that the approach which employs reference signal to do the ICA can de-noise the EEG signal effectively.In the end, under the limitation number of EEG signals, in order to extract the feature of EEG signals efficiently, put forward the method combined wavelet package with Independent Component Analysis. Firstly, according to the frequency order, decomposed the original EEG signals by using wavelet package transformation, and recomposed the related decomposed coefficients as a reference ofαrhythm. Then, take these series signals and EEG signal to be the input of ICA mixing matrix, the FastICA algorithm is adopted to separate the signals, thereby realize theαrhythm extracting from EEG signal, and analyze the power chart of the extractedαrhythm. The frequency ofαrhythm extracted concentrate between 8 Hz and 13Hz. It shows that the rhythm extracted is correct.