简介:Thecriticaltechnicalproblemofunderwaterbottomobjectdetectionisfoundingastablefeaturespaceforechosignalsclassification.Thepastliteraturesmorefocusonthecharacteristicsofobjectechoesinfeaturespaceandreverberationisonlytreatedasinterference.Inthispaper,reverberationisconsideredasakindofsignalwithsteadycharacteristic,andtheclusteringofreverberationinfrequencydiscretewavelettransform(FDWT)featurespaceisstudied.Inordertoextracttheidentifyinginformationofechosignals,featurecompressionandclusteranalysisareadoptedinthispaper,andthecriterionofseparabilitybetweenobjectechoesandreverberationisgiven.TheexperimentaldataprocessingresultsshowthatreverberationhassteadypatterninFDWTfeaturespacewhichdiffersfromthatofobjectechoes.Itisproventhatthereisseparabilitybetweenreverberationandobjectechoes.
简介:Inunderwatertargetdetection,thebottomreverberationhassomeofthesamepropertiesasthetargetecho,whichhasagreatimpactontheperformance.Itisessentialtostudythedifferencebetweentargetechoandreverberation.Inthispaperbasedontheuniqueadvantageofhumanlisteningabilityonobjectsdistinction,theGammatonefilteristakenastheauditorymodel.Inaddition,time-frequencyperceptionfeaturesandauditoryspectrafeaturesareextractedforactivesonartargetechoandbottomreverberationseparation.Thefeaturesoftheexperimentaldatahavegoodconcentrationcharacteristicsinthesameclassandhavealargeamountofdifferencesbetweendifferentclasses,whichshowsthatthismethodcaneffectivelydistinguishbetweenthetargetechoandreverberation.