EEG Signal Feature Extraction And 3-D Construction Based On WICA

Electroencephalograms(EEGs) are recordings of the electrical potentials produced by the brain. Analysis of EEG activity not only has been achieved principally in clinical settings to identify pathologies and applied to braincomputer-interface(BCI), but also has been given some of basic research in the field of the brain science. The purpose of EEG signal processing is to extract the weak EEG rhythm signals from the EEG signals with low SNR. We often used blind source separation(BSS) technology to extract some rhythm signals which controls thought and behavior. If we directly employ blind source separation techniques for multi-channel EEG signals processing, we may get a bad result or relatively large error. Time-frequency signal-processing algorithms such as discrete wavelet transform(DWT) analysis are necessary to address different behavior of the EEG in order to describe it in the time and frequency domain. We also propose a new EEG feature extraction method based on ICA algorithm combined with the wavelet transform method named WICA method.We get a good result.The traditional approaches for EEG signal processing are mostly on the time domain analysis and simple frequency filtering. When we apply them to address different behavior of the EEG, we can not get good results. Because of the nonlinearity, nonstationarity and strong randomness of EEG, the results obtained from those traditional approaches are not very satisfying. The wavelet transform is a time-frequency analysis method developed in the 1980 s, and has gradually become a useful tool for the analysis of EEG. It can not fundamentally solve the problem of signal and noise aliasing in the frequency domain, therefore, the improvement of the results of EEG signal processing is limited.Principal component analysis(PCA) and independent component analysis(ICA) are well known methods for multivariate data analysis. The results obtained from these two methods are not affected by the signal spectrum aliasing. PCA, which is aimed at Gaussian source, is derived through analyzing the second-order statistical properties of signals. But ICA is a higher-order statistic analysis tool, it can handle non-Gaussian signals. We generally believe that EEG is a non-Gaussian signals, so ICA on EEG signals has been a hot area of research in recent years.First of all, the basic knowledge of mathematical statistics and information theory is introduced. That is the basic knowledge of ICA. We use reference signals as the input of Fast ICA algorithm to remove the power line interference from the whisker barrel cortex local field potential(LFP) signals of mice. Secondly, we use wavelet transform theory to remove high frequency noise from LFP signals. And then, in order to extract the characteristic rhythm from LFP signals, we adopt the wavelet packet decomposition algorithm and reorganize the sub-band to get four basic rhythms. In this article, we discuss the four basic rhythm signals in frequency domain and in time domain. Next, we use the four basic rhythms as the reference signals of Fast ICA algorithm for further extraction, and we also get four basic rhythms. We also talk about the four basic rhythms which are obtained from Fast ICA algorithm in frequency domain and in time domain. We can see that WICA algorithm is more suitable than wavelet packet algorithm for EEG analysis, when we contrast the results obtained by the two methods. Finally, we arbitrarily take out three signals from the four basic rhythms and use Maltlab to obtain a 3-D graphics. It contributes to the exploration of medical diagnosis and brain-computer interface technology.