The Analysis And Identification Of Epileptic EEG Based On Chaos

Epilepsy is a common chronic neurological disease,caused by a sudden abnormalsynchronous firing neurons in the brain. It has become a major problem in themedical profession, seriously endanger human health. Currently, the EEG is one ofthe most important means of epilepsy diagnosis. But the brain belongs to a chaoticsystem, when doctors attempt to achieve the diagnosis of epilepsy, it need forlong-term observation of EEG in order to make the interpretation, it takes a lot of timeand effort and easy incorporation of subjective personal physicians so not tend to leadto accurate results. Therefore it is necessary to propose a computer-aided, the use ofEEG chaos characteristics, enabling the seizure signal classification method.Current methods of EEG analysis is divided into two categories: linear andnonlinear analysis method analysis. The linear analysis method is based on the linearcharacteristic of EEG, such as the relative amplitudes, relative energy etc., using timedomain analysis, frequency domain analysis and a method of time-frequency binding.Nonlinear method is based on nonlinear EEG characteristics, such as characteristicapproximate entropy, Lyapunov exponent, Long-range correlation and so on.Nonlinear dynamics of the brain as a system, with non-linear approach to the study ofEEG features is better. In this paper, the nonlinear characteristics approximate entropyand Lyapunov exponent scaling index to identify the epileptic EEG.This article also selected support vector machine as a classifier, its outstandingadvantages are particularly good performance in generalization,the ability to achieveerror is very small, and the robustness of the classification method is particularly good,extremely wide range of applications.This paper proposes two EEG classification methods, one is based on the classification and characterization methods Lyapunov exponent EEG amplitudefluctuation intensity fluctuations, one is based on approximate entropy andcharacterization of EEG amplitude of the normalized amplitude classification.Research using Epileptic data from the German born epilepsy research laboratory,using support vector machine as epileptic EEG classification. Lyapunovcharacterization of the EEG time series of the sensitivity of the initial system,approximate entropy to characterize the complexity of the EEG time series, theyrepresent the non-linear characteristic EEG; and volatility and fluctuations in thevalue of EEG amplitude normalized A value is portrayed its linear characteristics,combined with both full advantage of the characteristics of the EEG signal andimprove the accuracy of epileptic EEG classification algorithm. The simulationindicates that the algorithm has advantages of low computational complexity and highefficiency.