ThepaperaddressestheproblemoftargetrecognitionusingHigh-resolutionRadarRangeProfiles(HRRP).Anovelapproachoffeatureextractionanddimensionreductionbasedonextendedhighordercentralmomentsisproposedinordertoreducethedimensionofrangeprofiles.FeaturesextractedfromradarHRRPsarenormalizedandsmoothed,andthencomparativeanalysisofthesimilarapproachesisdone.Therangeprofilesareobtainedbystepfrequencytechniqueusingthetwo-dimensionalbackscattersdistributiondataoffourdifferentaircraftmodels.Thetemplatematchingmethodbynearestneighborrules,whichisbasedonthetheoryofkernelmethodsforpatternanalysis,isusedtoclassifyandidentifytherangeprofilesfromfourdifferentaircrafts.Numericalsimulationresultsshowthattheproposedapproachcanachievegoodperformanceofstability,shiftindependenceandhigherrecognitionrate.Itishelpfulforreal-timeidentificationandtheengineeringimplementsofautomatictargetrecognitionusingHRRP.Thenumberofrequiredtemplatescouldbereducedcon-siderablywhilemaintaininganequivalentrecognitionrate.
