Study of Epileptic Electroencephalogram Using Bispectrum Analysis
Li Changqiang, Liyu Huang, Fengchi Ju, Huang Yuangui, Cheng Jingzhi · Beijing shengwu yixue gongcheng · 2004
The bispectrum analysis method introduced here is quite effective to process the non-Guassian and nonlinear stochastic signals. So it is a useful tool for studying electroencephalogram (EEG) signal.The higher-order statistics methods were used to estimate the bispectrum of the Epileptic EEG at different phases.The estimation revealed some new information about Epileptic EEG, which could not be obtained by the conventional power spectrum methods. The experimental results indicated the Epileptic EEG of different phase had different deviations from Guassian signal, where the bicoherence coefficient could distinguish signal features of EEGs at different stages of acting up.The bicoherence index, therefore, could be a useful parameter for clinical monitoring and for prediction of acting up.