摘要
InordertoexplorethecorrelationbetweentheadjacentsegmentsofalongtermEEG,animprovedprincipalcomponentanalysis(PCA)methodbasedonmutualinformationalgorithmisproposed.Aone-dimensionEEGtimeseriesisdividedequallyintomanysegments,sothateachsegmentcanberegardedasanindependentvariablesandmulti-segmentedEEGcanbeexpressedasadatamatrix.Then,wesubstitutemutualinformationmatrixforcovariancematrixinPCAandconducttherelevanceanalysisofsegmentedEEG.Theexperimentalresultsshowthatthecontributionrateoffirstprincipalcomponent(FPC)ofsegmentedEEGismorelargerthanothers,whichcaneffectivelyreflectthedifferenceofepilepticEEGandnormalEEGwiththechangeofsegmentnumber.Inaddition,theevolutionofFPCconducetoidentifythetime-segmentlocationsofabnormaldynamicprocessesofbrainactivities,theseconclusionsarehelpfulfortheclinicalanalysisofEEG.
出版日期
2013年03月13日(中国期刊网平台首次上网日期,不代表论文的发表时间)