Research Of The Method Of Fault Feature Extraction Of A Class Of Complex System Based On EEG Analysis

The brain is the most complex and intelligent dynamic information processing systemas we known. In a sense, the brain can be regarded as a dynamic network whichcontinuous remodeling its function connection. As the complex network theory andnonlinear dynamics method becoming a hot spot of interdisciplinary research, it provides anew path for us to study the brain–a complex giant system. We study the brain as acomplex system fault output characteristics. Then we can forecast and treatment somebrain diseases. From this perspective, the study of the brain also can simulate artificialintelligence system, and it can help the fault diagnosis of complicated systems. Therefore,the study of the brain electrical signal characteristics extraction method has importantsignificance and application prospect.In this paper we analyze the EEG signals from two aspects. They are the nonlineardynamic method and complex network theory. We use two kinds of nonlinear dynamicsanalysis methods to analyze EEG signals, and they are recursive analysis method andmultivariate graph centrobaric trajectory method. Recursive analysis method puts timeseries in the recursive diagram, and it figures the characteristics of signals and inherentlaw directly from recurrence plot. Firstly, the multivariate graph centrobaric trajectorymethod reconstructs phase space of the signal. Then it finds the centrobaric trajectory inthe phase space of multivariate graph. We analyze the characteristics of the centrobarictrajectory in order to observe the different feature of the signal. Complex network analysisis to establish the network topology. Through observing the topology of the network, wecan find the signal characteristics.Through analysis the recurrence plot of EEG signals, we found that the EEG signalsduring disease is chaotic, and normal EEG signals not. We extract multivariate graphcentrobaric trajectory torque characteristics of the EEG signals to distinguish between theabnormal EEG signals and normal EEG signals. Through constructing the complexnetworks of EEG signals we can observe the dynamics characteristics of EEG signals. Ithas very good effect to identify the epileptic seizure and the normal period of EEG signals.Epilepsy signal feature extraction provides certain theoretical basis for the epilepsydiagnosis and forecast. At the same time, it provides new way for the complex systemsfault diagnosis method.