简介：

简介：Shannonentropyintimedomainisameasureofsignalorsystemuncertainty.Whenbasedonspectrumentropy,Shannonentropycanbetakenasameasureofsignalorsystemcomplexity.Therefore,waveletanalysisbasedonwaveletentropymeasurecansignifythecomplexityofnon-steadysignalorsysteminbothtimeandfrequencydomain.Inthispaper,inordertomeettherequirementsofpost-analysisonabundantwavelettransformresultdataandtheneedofinformationmergence,thebasicdefinitionofwaveletentropymeasureisproposed,correspondingalgorithmsofseveralwaveletentropies,suchaswaveletaverageentropy,wavelettime-frequencyentropy,waveletdistanceentropy,etc.areputforward,andthephysicalmeaningsoftheseentropiesareanalyzedaswell.TheapplicationprincipleofwaveletentropymeasureinElectroEncephaloGraphy(EEG)signalanalysis,mechanicalfaultdiagnosis,faultdetectionandclassificationinpowersystemareanalyzed.Finally,takethetransmissionlinefaultdetectioninpowersystemforexample,simulationsintwodifferentsystems,a10kVautomaticblockingandcontinuouspowertransmissionlineanda500kVExtraHighVoltage(EHV)transmissionline,arecarriedout,andthetwomethods,waveletentropyandwaveletmodulusmaxima,arecompared,theresultsshowfeasibilityandapplicationprospectofthesixwaveletentro-pies.

简介：Anewapproachtoknowledgeacquisitioninincompleteinformationsystemwithfuzzydecisionsisproposed.Insuchincompleteinformationsystem,theuniverseofdiscourseisclassifiedbythemaximaltoleranceclasses,andfuzzyapproximationsaredefinedbasedonthem.Threetypesofrelativereductsofmaximaltoleranceclassesarethenproposed,andthreetypesoffuzzydecisionrulesbasedontheproposedattributedescriptionaredefined.ThejudgmenttheoremsandapproximationdiscernibilityfunctionswithrespecttothemarepresentedtocomputetherelativereductbyusingBooleanreasoningtechniques,fromwhichwecanderiveoptimalfuzzydecisionrulesfromthesystems.Atlast,threetypesofrelativereductsofthesystemandtheircomputingmethodsaregiven.

简介：Duetothefeaturesofthemulti-spectralimages,theresultwiththeusualmethodsbasedonthesupportvectormachine(SVM)andbinarytreeisnotsatisfactory.Inthispaper,afuzzySVMmulti-classclassifierwiththebinarytreeisproposedfortheclassificationofmulti-spectralimages.Theexperimentisconductedonamulti-spectralimagewith6bandswhichcontainsthreeclassesofterrains.Theexperimentalresultsshowthatthismethodcanimprovethesegmentationaccuracy.