简介:Usingtheextremedifferenceofself-similarityandkurtosisatlargelevelscaleofwavelettransformapproximationbetweenthePTFM(PulseTrainsofFrequencyModulated)signalsanditsreverberation,afeature-basedmatchedfiltermethodusingtheclassify-before-detectparagriamisproposedtoimprovethedetectionperformanceinreverberationandmultipathenvironments.Processingthedataoflake-trailsshowedthattheprocessinggainoftheproposedmethodisbiggerthanthatofmatchedfilterabout10dB.Inmultipathenvironments,detectionperformanceofmatchedfilterbecomebadlypoorer,whilethatoftheproposedmethodisimprovedbetter.Itshowsthatthemethodismuchmorerobustwiththeeffectofmultipath.
简介:
简介:Thenumericalquadraturemethodsfordealingwiththeproblemsofsingularandnear-singularintegralscausedbyBurton-Millermethodareproposed,bywhichtheconventionalandfastmultipoleBEMs(boundaryelementmethods)for3Dacousticproblemsbasedonconstantelementsareimproved.Tosolvetheproblemofsingularintegrals,aHadamardfinite-partintegralmethodispresented,whichisasimplifiedcombinationofthemethodsproposedbyKirkupandWolf.Theproblemofnear-singularintegralsisovercomebythesimplemethodofpolartransformationandthemorecomplexmethodofPART(ProjectionandAngular&RadialTransformation).Theeffectivenessofthesemethodsforsolvingthesingularandnear-singularproblemsisvalidatedthroughcomparingwiththeresultscomputedbytheanalyticalmethodand/orthecommercialsoftwareLMSVirtual.Lab.Inaddition,theinfluenceofthenear-singularintegralproblemonthecomputationalprecisionsisanalyzedbycomputingtheerrorsrelativetotheexactsolution.ThecomputationalcomplexitiesoftheconventionalandfastmultipoleBEMareanalyzedandcomparedthroughnumericalcomputations.Alarge-scaleacousticscatteringproblem,whosedegreeoffreedomsisabout340,000,isimplementedsuccessfully.Theresultsshowthat,thenearsingularityisprimarilyintroducedbythehyper-singularkernel,andhasgreatinfluencesontheprecisionofthesolution.TheprecisionoffastmultipoleBEMisthesameasconventionalBEM,butthecomputationalcomplexitiesaremuchlower.