《中国邮电高校学报(英文)》论文投稿模板.docx

《《中国邮电高校学报(英文)》论文投稿模板.docx》由会员分享，可在线阅读，更多相关《《中国邮电高校学报(英文)》论文投稿模板.docx（10页珍藏版）》请在课桌文档上搜索。

1、Noisyspeechemotionrecognitionusingsamp1.ereconstructionandmu1.tip1.e-kerne1.1.earningJiangXiaoqing121XiaKcwcn1(0),1.inYong1.iang,BaiJianchuan1. Schoo1.XE1.cctnx2Infixnuiitxi1.Acrn.IfcbeiUfIiYCrtiI)MTbtKKihgy.TianjinXOU1.I.Chuu2. Schoo1.ofInZormutionScuetuxandEngiCKCnc:,UnivCnjCyfJirun.J1.n由25K22.Cu3

2、. InfoEXiiionCeneer.HanjinChsiian1.ivcnMy.TiCejin5(MMH4.ChinaAbstractSpeechemotionrecognition(SER)innoisyenvixmmen1.isavita1.issueinartificia1.ime1.Iigence(A1.).Inthispaper,(hereconstructionofspeechsamp1.esrcnu)vcstheaddednoise.Acousticfeaturesextractedfromthereconstructedsamp1.esarcse1.x1.cd(obui1.

3、danopirna1.feuresubsetwithbe1.terCmO1.iona1.11xognizabih1.y.Amuhipk-keme1.(MK)support,ectormachine(SVM)c1.assifierso1.vedbyscni-dc11nitcprogramming(SDPIisadoptedinSERprocedure.TheproposedInC1.hodinthispaperisde11ms1.rbus1.whennmseexists.KcywHdxv1.MrtionsisneededtomakeaPrOperresponse.Inthispaper,nois

4、ySERisstudiedusing(hecombinationofsamp1.ereeouirUC1.ionbasedoncompressedsensing(CS)theoryandmu1.tip1.eke11e1.1.earning(MK1.).InSER1.woessentia1.aspectsinf1.uencingthePCrfOfmansoftheemotionrecognitionsystemW2OI6CxespriirJutk1.r：XuiKCQCr1.Emai1.:kw*i心MbinVdUeDOI:10.1016SI(K588851.17*H*features,andattc

5、mpposedbyDonohoe(a1.providespromisingmehcKtoOOiSyspeechprocessing6-7.SparsercpfsenatininCStheoryhasbeenusedinnonarane(ricc1.assifier.Zhaoe(a1.adopted1.heenhancedsparseNPZMm1.i1.1.iOac1.assifierIodeus1.SER.Addi1.iuna1.1.y.asIhcderivedCocffiden1.sofnisearersparseinanytransferd(xnain.itisimpossib1.eIor

6、ve(heI1.cxibi1.i1.yofkerne1.fur(i.MK1.isProPOMrdanddeveked(bina(knofdif1.crcntkerne1.s.1.anCkriCtcta1.PrOPOSCdMK1.withatransductionsettingfor1.earningakerne1.matrixfromdata.Themethodaimedattheopibinationofpredefinedbasekerne1.stogenerateagoodtargetkerne1.(1.Jincta1.Pft)POScdfeaturefusionmethodbasedo

7、nMK1.toimprovethetoa1.SERperformanceofc1.eansamp1.es.TheWdghISofdi11crcnkerne1.scorrespondingtotheghba1.and1.oca1.featuresaregivenhughagfidSearChnhve(heSVMnde1.inabinaryInjeMructuredmuki-c1.assc1.assifier,and(hefusionCOefnCien1.qOfdifferentkerne1.sareso1.vedbyIhcSDP(ofindUPIima1.WeighISofmuI1.ip1.ck

8、erne1.s.The3gofthep;iprarcstrc1.uredasIhcfo1.1.owings:Sect2reviews1.hebasicideaofCSinspexhsigna1.processingandana1.yzestheperformanceofnoisysamp1.ereconstruction.Sect.3introducesMK1.so1.vedbySDRAcousticfeaturesandfeaturese1.ectionarcpresentedinSect.4.ThCpcribrmanceeva1.uationofSERandexperimenta1.res

9、u1.tsarci1.1.ustratedandana1.yzedinSect.5.Fina1.1.ySect.6devotestotheconc1.usions.2 CSandsamp1.ereconstructionofnoisyspeechCScombinessamp1.ingandcomssiinintooMepusingIbeminimumnumberofmeasurementsWhhmaximuminfonnation.CSaimsIurecoversparseMgna1.withfarIg(hanNyquis1.*Sh;mnonsamp1.ingr*te.u11dIberecon

10、structioncanbeexactunderkeyConCCPIysuch-1.InEq.(I)Vzisbeorthogona1.basisIM1.rix,a1.sonamedrepresentationmatrix,=(*.，“)isprojectioncoefficient,isIhenecioncoefficientInaIriXand=V,1x.Itcanbesaidthatxandaarctheequiva1.emrepreenuwnsofIhesamesigna1.withxintimedomainwhi1.eaindomain.Whenthesigna1.xon1.yhask

11、non-zerorcocftkicn(sandkN.nisIhCSPaZtsisofxandxcanbeconsideredksparsewih*rserepresentaionofEq.(1).InCS1.ben.IhesensingprocessCanberepresentedas:尸极(2)InEq.(2)isthejVasurcmcntmatrix,andFE(MVVMisthemeasuremenvectorof-dimensina1.Compressionisrea1.izedbecause(hedimensionofneasurenntsyisfar1.essthan(hedim

12、ensionofthesigna1.x.WithEq.(I),Eq.(2)canberewrienas:”6M(3)where0=isMNdimensiona1.recmMruc1.ionmatrix,andaisksparsevectorrepresenting1.heprojectioncoefficientsofxinVxdomain.Reconstniciiona1.gorithmsinCSIryIOSO1.veEq.(3).whichisanIinderde1.ermivdequationwithoutadeterminantso1.ution.Whenthesigna1.isspa

13、rseandsatisfiestheRIPcondition.asparseapproximationso1.ution(OEq.(3)canbeobtainedbyminimizing(he1.1-norm.RIPofna1.rixisdefinedoniisoine1.ryconMan(c(O.I)fora&sparsesigna1.xani1.Miiisfies:1.-!1.+mmatrixarenear1.yorthogona1.Theequiva1.enconditionofRIPis(heiIKobereiKCbetweenIiwaNurenicntmatrix;ind(herep

14、resentationmatrix.Avarietyofreconstructionmethodssuchasgreedya1.gorithmsandconvexQP1.imi/a1.iencanbeusedinIhCso1.vingPnKofEq.(3)B-2U.WhenCStheoryisapp1.iedtospeechsigna1.processing,theprerequisiteistoachievethesparserepresentationofspeechsigna1.susingproperorthogona1.basis.ThCexcitationofvoicedandun

15、voicedspeechisquasi-periodicvibrationsofvoca1.cordsandrandomnoiserespective1.y.Sovoicedspeechcarriesthemostenergyofthesamp1.eandfocusesin1.owerfrequencyMXtion.OneofthemostimportantspecmCkiracteristicsofdiscretecosinetransformaion(DCT)is(he3r11ngenergyconcenraionin1.owfrequencycoetticents,whichmakess

16、uitab1.etoana1.yzethesparsityofspeechsigna1.s.T1.ieorthDCTCoefYicieniafn)ofaspeechframexWiIhNsamp1.escanbeca1.cu1.atedby:*、11(2-X-1.)IC.,a(n)=h(w)2-()cos-：/n=1.2NST2Af-J-：w三IHm)-E后2这QwhereW“)denotesthe爪hsamp1.eofthespeechframe.Examp1.esofc1.eanvoicedframeandunvoicedfraneswe1.1.asIheirDCTcoefficients

17、arep1.ottedinFig.1.Obvious1.y,on1.yafcv-DCcoefficientshave1.argeramp1.itudewhi1.e(herestsarctowa111.szero.Thesparsityismoreobviousinthevoicedframe.ThereforetheDCTcoefficientsofvoicedSPCNhsigna1.scanbeconsideredasksparseapproximate1.y.2aB$01.ISO3002S0300350NumberDCTedIiMn1.xManunvuicedrajurbDCTc)tfff

18、kicnufa,ccdundUnV(Hadframe%AccordingtoCStheory,voicedsigna1.scontaminatedbynoisecanbereconstructedwithhighqua1.ity.Fig.2p1.otsthedenoisingperformanceofsamp1.ereconstruction.IfrandomGaussianmatrixisusedasthemeasurementmatrix,compressivesamp1.ingmatchingpursuit(CoSaMP)14.orthogona1.matchingpursuit(OMP

19、116.basispursuit(BP)I7andPO1.ytoPCfacespursuit(PFP)21a1.gorithmsarcadoptedinthereconstructionofnoisysamp1.es.Thenoisyvoicedframeisproducedbyadded2()dBGaussianwhitenoiw:onthec1.eanframe.Itisc1.earthatIberoco11sc(edsamp1.eshaveapproximatequa1.ityothec1.eanwaveform.Theamp1.itudeand(hePeriOdofthec1.eanf

20、ramearepreservedin(hereeonsnc(edframes.Tereconsmc(kinwaveformsofBP,OMPandPFPa1.noscoincide.Thehighqua1.ityofthereconstructedvoicedspeechensures(heprecisionof(hefeausUX1.nK1.iQninSER.BO1.hFig.1andFig.2denvons(1111ebesamp1.ereconshuc(nisfeasib1.einnoisySER.06(4Y)6Rcxms1.nK1.cd.BPRccvrrdrw1.atPFPfRwmzM

21、OMPRcdy1.isthec1.ass1.abe1.ofx1.ThCninthefeaturespaceinducedbymappingfunctione.wecanfindahypcrp1.ancwiththemaximummargintoc1.assifytwoc1.asseswithdiscriminantfunction:/(=*.*)frwherewandbarcweightvectorandtheOifSCt(ha(canbecomputedbyso1.vingaquadra1.k:OPIimiZa1.ionprob1.em:.Inun-mh“2“依XJ*bN1.;三1.2.JT

22、omakethemethodmoref1.exib1.eandrobust,uhypcrp1.anccanbeconstructedbyre1.axingconstrainsinEq.(7).which1.eadstothefo1.1.owingsoftma生informu1.ationwiththeintroductionofs1.ackvariab1.es1.toaccountfbrrnisckisiGcaUons.Thebjec1.ivefunctionandconstraintscanbeformu1.atedas:Imin*m,+CV&2yc,(M(x,)+ft)1.-;0.I=1.

23、Z./where/isthenumberof(rainingpatterns.CisaparameterWhkrhgivesatradeoffbetweenmaximummarginandc1.assificationerror,andisamappingfromtheinputspacetothefeaturespace.Eq.canbeso1.vedbyintroducing1.agrangcmu1.tip1.iers:UW4b.m=gM+：-SSd-i-1i-1(ZaJM(WI幽与)+为T+帛(9u1.wherea1.0and力R,i=12Jare(heUigrangenw1.(ip)i

24、er.Byse1.1.ingPar1.Mderivativesof1.(ozeroyndsubstituting(hesu1.sintoEq.(9).w.1.canbee1.iminatedandEq.(9)canbetransformedfo1.1.owingWo1.fedua1.form:max”,-Hy,y,a1.atk(x1.xt)U1.*.：S.t.Caj0.ZKa,=S=1,2Iwherek(xt.xy)=(x1.).(xj)isakerne1.Eq.(IO)Canberewri11eninmatrixfbmas:maxie-,G(K-2s.t.0Ce-0arJ=0wheree三1

25、.I1.jandGK)=diag(y)KdugOr).diag(j)intothe(IO)function.（三）isthediagona1.matrixwithdiagona1.yandKisthekerne1.matrixwithK”=A(x,.xJ.f=1.2,.,/,j=1.2./.3.2MKSVMThePerfbnniinWofaSKn1.hoddependsheavi1.yonIhechokeofIhekerne1.Kerne1.fusionhasbeenPrvPoMXJtodea1.withthisprob1.emthrough1.earningakerne1.machinewi

26、thMKsIO.22.Oneoftheeftc(ivckerne1.fusionstrategicsisaweightedcombinationofmu1.tip1.ekerne1.s.ThCcombinedkerne1.fu1.inis(x,.x,)=EXA.x力,Wherex)=(jr),1*,(j:)内,(*)andMisthenumberofkerne1.func(ionstobecombined.Thecorrespondingkerne1.nirixcanbewrittenas:K=N,(12)whereK1.1.wSWiwisthekerne1.matrixconstructed

27、.ufrom,andM(*1)isIhCre1.atingweight.1.anckrictcta1.proposedaMK1.methodwithaIransductionsettingtoobtain1.AccordingtoEq.(II).trainingtheSVMfor;igivenkerne1.invo1.vesyie1.ding(heoptima1.v1.ucoftKy-maxae-,G1.K),whichis2afunctionofIbCparticu1.archoiceofthekerne1.matrixobvious1.y.Sofindingthekerne1.matrix

28、canbeconsideredasanOPIin1.i/a1.ienprobkmIha1.means(oI1.ndKinsonicconvexsubsetofpositivesemi-definitematriceskeepingthetraceofKconstant:min0trK=cG(K1.0Abinantreestructurei1.1.ustratedinFig.3isadoptedinthestructuredesignofthemu1.ti-rest,orhierarchySVMstructures(2-25.In(hebinary【攸sruc(uv.(hefirstc1.ass

29、ifyingnode(Modeh)isimprovedbyMKSVMICrecognize(hemostconAisabkenx)inwhi1.e(he(Ieeperc1.assifyingnodes(Modek-M(Xfeh)s1.i1.1.retainSKSVM.TakeIhcBer1.inDatabaseofEno(ma1.Speech26furexamp1.e,happyis1.hcmainfactorinf1.uencingovera1.1.PCdbEUMKCofIhec1.assifierbcciuseofits1.owestrecognitionaccuracy24,27-28.

30、1.hus.Mode1.canbeused(orecognizehappyirmotheremotionswhenBer1.inDatabaseofEmotiona1.Speechisst1.icd.Thisarrangementcanreduceerroraccumu1.ationcausedbythemostCputingcomp1.exityofso1.vingSDPinCvcrymode1.PUp怨ngfcirMUtrJ1.3I卜灯。IIAnzCraI3RneOIFmE血adF.3s1.nc1.三eMuIii-CkiKSandMKC1.wifkrwithbinrytree4Acoust

31、icfeaturesandfeaturese1.ectionFe;I1.Urese1.ectionisnevewniryinbui1.dingOPei1.na1.fea(urcSUbMJ1.withemotiona1.rccognizabi1.i1.y.DoUb1.Cinputsymmetrica1.re1.evance(DISR)isaninform;Uiontheoreticse1.ectioncriteriondependingontheuti1.izationofsymmetricre1.evancetoconsiderthecomp1.ementaritySpeechfeatures

32、usua1.1.yusedinSERarc1.hcprosodicfeatures,voicequa1.ityfeaturesandspectra1.features.Pitch,energy,duration,formants,andmc1.frequencyccpst11mcoefficients(MFCC)andtheirstatisticsparametersarcextractedinthispaper.Thetota1.dinnsionofthefeaturevectoris45.Tab1.e11.iststheacousticfeaturesadoptedinthefo1.1.o

33、wingexperiments.Tab1.e1AeUxHCfeaturesIyPCFeMvrvSuiiMicparameter%PiichMkiwwm.mimum,“2c11nf.hr*rdOcvitfkak11f*1.qiurti1.c.niei1.ian.I1.iifd*rti1.c.rt-quM1.iru”趾PrvwdicEnEIyMaxmm.mnmum.rarc.mcun.1.ar*rd改ViMjC.IirSIqtunik.median.IhirdQJdniIc.irrquMikranpc仙UreOmmimTHa1.mk*ivkvMFVCI2mjcfMFCCbetweenIWOinputfeatures.TeIMinadvantageofDISRCnieriixiis(hese1.eccedCOmPIUInefHaryvariaNehasmuchhigherrbabi1.iyre1.evanceona1.1.Of1.hedoub1.einputsin(hesubset29.5 Experimenta1.resu1.tsandana1.ysisIn(hispaperIheBer1.inDatabaseofEmoiiona1.Speechisse1.ected(ogtheproposedmethodfromIheaspectofs