| Cerebrum is the most complicated and important organo of the human body, because it is the main part of the thought and the centre of the nerves activity. EEG signals collected from cerebral cortex by electrodes contain abundant information about cranial nervous activities. EEG is a kind of continuous electric pulse created by the stimulations of numerous cerebrocells. Because EEG synthetically characterized the functions and states of cerebral activities, it has become a mainstream to analyse EEG with all kinds of methods. However, its research exists many problems, as EEG is very complicated and particular.It is preliminarily confirmed and accepted by academe that EEG is chaotic and possess many nonlinear dynamic characteristics, so we can analyse EEG with nonlinear dynamic methods. After its nonlinear dynamic characteristics are fleetly extracted, we can seek the relationship between those parameters and cerebracal states or diseases. The results can be used to estimate the cerebration and to monitor the healing process of encephalopathy.In this paper, based on the nonlinear dynamics and the investigation on the EEG research at home and abroad, we develop and discuss the nonlinear dynamics analysis methods. The main works and conclusions are as follows:(1) After the studies on nonlinear dynamics, chaos and chaotic characteristics, the chaotic properties of EEG are discussed and proved by experimental methods.(2) It is the base and precondition that EEG must be preprocessed before it is available. In this paper, in order to reduce the influence possibly created by interferences and noises, a series of digital filters are used to reduce the 50Hz line-frequency disturbs and baseline drifts, and the theory of bursts of 40Hz EEG are used to get grid of the possible disturbances of EMG. So the stability and reliability of the results are greatly improved.(3) Traditional G-P algorithm based on phase-space reconstruction and used to calculate the correlation dimension of time series is improved in this paper. The factors, including embedding dimension, time delay, data length, scale of linear regression and amplitude of EEG signals, which possibly affect the results, are discussed. At last, reference standards to select those parameters are given.(4) The algorithm of calculating largest Lyapunov exponent introduced by A.Wolf and based on evolution theory has many defects, such as great calculation capacity and weak antijamming ability. By comparison, we discove that the small data algorithm presented by M.T.Rosenstrin is more ascendant and is applied in this paper to estimate largest Lyapunov exponent. (5) As complexity measures of time series, approximate entropy and Kolmogorov complexity possess many merits, such as small calculation capacity, simple algorithm and strong antijamming. They are used to characterize the complexity measure of EEG. From the clinical results, they are also good parameters to distinguish the different states and samples.(6) The raw EEG data under different thought states containing 30 normal objects and 26 patient objects are collected, and 20 normal and patient samples are respectively selected as effective experiment objects. For experiment objects, all dynamical parameters containing correlation dimension, Kolmogorov complexity, Lyapunov exponent and approximate entropy are calculated to validate the correctness of the methods provided in this paper. After the calculation results are contrasted and analysed, conclusions are presented as follows:The correlation dimension under calm states is clearly less than under thought states, and under simple thought states is less than it under complicated thought states. The of patient is visibly less than normal sample too. But the of patients under different thought states can not be distinguished clearly. Kolmogorov complexity and approximate entropy have similar contrasting conclusions as and Lyapunov exponent has inverse contrasting conclusions as . For all parameters, under different conditions, there is ob...
|
PDF Full Text Request