资源描述:
《5 Gait authentication based on acceleration signals of ankle》由会员上传分享,免费在线阅读,更多相关内容在学术论文-天天文库。
1、ChineseJournalofElectronicsVol.20,No.3,July2011GaitAuthenticationBasedonAccelerationSignalsofAnkle∗LIYuexiang1,WANGXiaobo1andQIAOFeng2(1.SchoolofComputerandInformationTechnology,ShanxiUniversity,Taiyuan030006,China)(2.FacultyofInformationandControlEngineering,Sheny
2、angJianzhuUniversity,Shenyang110168,China)Abstract—Anovelfeaturepointsbasedgaitauthen-WS-basedgaitrecognitiontechnologyisrelativelynew,es-ticationmethodisintroducedinthispaper,whichusespeciallywiththedevelopmentofMEMStechnology;sensorstheaccelerationsignalsacquired
3、fromanankle-mountedareincreasinglyminiaturizedandintegrated,andhavebeen3-axisaccelerometer.Featurepointsareextractedfrom[3][4]widelyusedingesturerecognition,kinematicsanalysisetc.originalgaitsamples.AdynamictimewrappingalgorithmInthismethod,gaitsignalsarerecordedby
4、wearablesensorsisemployedtomatchthefeaturepointsofdifferentsamplesandcalculatethedistortions.Basedonthesedistortions,aandthenutilizedforindividualidentification.multi-criterionmodelisdesignedforauthentication.TheAilistoetal.firstproposedthegaitauthenticationmethodexpe
5、rimentalresultshowsthattheextractedfeaturepointsbasedonaccelerationsignals[5].Theyusedawaist-mountedcanrepresenttheoriginalsignalsgoodenoughinauthenti-accelerometertocollectgaitdata.Theirmethodisfirstlyfind-catingwithoneaccelerometer,andtheequalerrorrateofingindividu
6、alsteps,normalizingandaveragingthem.Thenthismethodisonly3.27%,betterthanthatoftheprevioustheeachstepisalignedwiththetemplateandcross-correlationliteraturesreported.iscalculated,whichisusedasameasurementofsimilarity.Keywords—Gaitauthentication,MEMSaccelerome-AnEqual
7、errorrate(EER)of6.4%isachievedintentativeter,Featurepoints,Dynamictimewarping,Multi-criterionexperimentsof36testsubjectsfinally.Besides,therearemodel.manystudiesbasedonseveralothermethods,suchasfre-quencydomainanalyze,distributionstatisticsandK-nearest[6−8]I.Introdu
8、ctionneighbor,andtheresults(measuredinEER)arebetween5.6%and19%.Biometricsystems,suchasfingerprint,face,andvoiceHuangetal.builtapairofintelligentsh