The Complexity Of The Eeg Signal Analysis

The complexity analysis is extremely important in biological signal processing.In this paper, the complexity analytical method was used to study electroencephalogram signal. Firstly, we should preprocess EEG. Then wavelet entropy, power spectrum, bispectrum and correlation dimension were used to analytical study the EEG signal. Finally, we summed up the complexity rule of the EEG. The study had five parts:The first part, We used wavelet packet transform to remove EEG series with low-frequency. Then a nonlinear filter was used to remove ectopic values from EEG So we achieved the goal of preprocessing EEG The results showed that the sharp difference between the two EEG subseries polluted by noise diminished after preprocessing. The LZC of two EEG subsequences chosen from the same EEG sequence polluted by noise is 0.2232±0.0730,0.4721±0.0705 before preprocessing. The LZC of the two EEG subsequences is 0.6286±0.0240,0.5911±0.0583 after preprocessing. The preprocessing denoised EEG and removed ectopic values form EEG effectively. The results demonstrated the necessity of preprocessing EEG. The preprocessing denoised EEG series effectively as well as reserved useful information in EEG preferably.From the second part to the fourth part, the wavelet packet transform was used to extract EEG in which there are five kinds of EEG rhythms. Then wavelet entropy, power spectrum, bispectrum and correlation dimension had been calculated. The statistical value revealed that:the complexity and its variation of brain were maximal in the subject's awake hour, were decreasing with the deepening of sleep, but the complexity in REM became larger than the complexity in deeper sleep. The four methods, from different point of view, had taken on respective rules about EEG complexity changing. Wavelet packet entropy had the advantage of calculating simply, steady result and representing preferably different sleep stages. The relative value of correlation dimension could effectively distinguish different sleep stages. The fifth part, we summed up the complexity rule which was derived from the second part till the fourth part. According to the state of the sleep, the wavelet entropy and the correlation dimension had taken on the similar variable rises or falls. The correlation coefficient between them proved our hypothesis.The work we had done just is the initial attempt, for the complexity analytical application in biology otherwise signals, we hope to have further study.