简介：Event-relatedpotentials(ERP)isanimportanttypeofbraindynamicsinhumancognitionresearch.However,ERPisoftensubmergedbythespontaneousbrainactivityEEG,foritsrelativelytinyscale.Furthermore,thebrainactivitiescollectedfromscalpelectrodesareofteninevitablycontaminatedbyseveralkindsofartifacts,suchasblinks,eyemovements,musclenoiseandpowerlineinterference.Anewapproachtocorrectthesedisturbancesispresentedusingindependentcomponentanalysis(ICA).ThistechniquecaneffectivelydetectandextractERPcomponentsfromthemeasuredelectrodesrecordingseveniftheyareheavilycontaminated.Theresultscomparefavorablytothoseobtainedbyparametricmodeling.Besides,auto--adaptiveprojectionofdecomposedresultstoERPcomponentswasalsogiven.Throughexperiments,ICAprovestobehighlycapableofERPextractionandS/Nratioimproving.