Eeg Signal Research Based On Partical Directed Coherence And Its Directed Network

Electroencephalogram(EEG)is one of the best ways to understand how the brain transmits information.Studying the directional network of EEG signals in multiple channels is important for understanding how different brain regions transmit information and understanding the causes of brain disorders.In this paper,a de-noising method based on Noise Assisted Multivariate Empirical Mode Decomposition(NA-MEMD)was studied,and the directional brain network of the de-noised EEG was studied.Firstly,the NA-MEMD algorithm is used to decompose the simulation signals of multiple channels into multiple(Intrinsic Mode Function,IMF)components.Then,through the Normalized Weighted sort of Mutual Information(Normalized Weighted Permutation Mutual Information,NWPMI)algorithm to calculate the simulation signals of each channel with the IMF component,introduced by the channel noise signal and the IMF component,the IMF component simulation signal and noise correlation between the IMF component,then calculate effective factor of each IMF component,to complete the IMF component screening.The selected IMF components were added to complete the signal reconstruction.In this paper,the mixed sinusoidal signal of multiple channels and the electroencephalogram signal composed by Jansen model are simulated and analyzed respectively,and the reliability of the reconstructed signal method is verified.Secondly,two parameter recognition algorithms for Multi Variable auto-regressive(MVAR)models are studied,Nuttall Strand(NS)algorithm and L1 norm algorithm.A partial Directed Coherence(PDC)algorithm based on MVAR is presented.Through simulation,the performance of the two algorithms is compared in many aspects,including:the influence of PDC calculated under different coupling coefficients on the synchronization direction between other channel signals,the influence of different channel number on the synchronization direction between channels,and the influence of different noise on the synchronization direction between channels.Simulation results show that the PDC algorithm based on NS is more stable than the PDC algorithm based on L1 norm,and it is applied to the EEG of Mild Cognitive Impairment(MCI)patients.Finally,the index of brain network such as degree,input degree,output degree and local transmission efficiency in complex network theory was studied,and the directed connection matrix between MCI and normal EEG in different frequency bands was analyzed.It was found that there were significant differences between the output degree in Beta band and the local efficiency in Gamma band,which were correlated with the score of neuropsychological scale.