EEG-signal-processing-脑电信号处理方法算法VIP免费

EEGSIGNALPROCESSINGEEGsignalmodelling1Availablefeatures2Classificationalgorithms3IndependentComponentAnalysis4SparseRepresentation51Bioelectricity1Signalgenerationsystem2ExcitationmodelLinearModelNonlinearModel2Basicfeatures1Modernmethods2TemporalAnalysisSignalSegmentation:labeltheEEGsignalsbysegmentsofsimilarcharacteristics.TemporalCriteriaFrequencyAnalysisSuboptimalDFT,DCT,DWT;OptimalKLT(Karhunen-Loève)Demerits:completestatisticalinformation,nofastcalculation.SignalParameterEstimationARmodel:Merits:OutperformDFTinfrequencyaccuracy.Demerits:sufferfrompoorestimationofparameters.Improvements:accurateorder&coefficients.ARcoefficientsestimationmethodsYule-Walkeraryule(x,p)Merits:ToeplitzmatrixLevinson-Durbin,fastest!!!Demerits:withwindowbadresolutionofPSDARcoefficientsestimationmethodsCovariancemethodarcov(x,p),armcov(x,p)Merits:withoutwindowgoodresolutionofPSDDemerits:slowBurgarburg(x,p)Merits:accurateapproximationofPSDDemerits:lineskewing&splittingComparisonPrincipalComponentAnalysisUsesameconceptasSVDDecomposedataintouncorrelatedorthogonalcomponentsAutocorrelationmatrixisdiagonalizedEacheigenvectorrepresentsaprincipalcomponentApplicationdecomposition,classification,filtering,denoising,whitening.3SparseApproximation1SparseDecomposition2Over-completedictionaryatomsHilbertspace:Signal:Error:“Sparse”:l<