您好,欢迎访问三七文档
当前位置:首页 > 商业/管理/HR > 质量控制/管理 > 基于到达时间差的声源定位算法与实现-叶诚
300384....TP391Adoi10.3969/j.issn.1673-095X.2015.02.012SoundsourcelocalizationalgorithmanditsimplementationbasedontimedifferenceofarrivalYEChengYANGShu-yingSchoolofComputerandCommunicationEngineeringTianjinUniversityofTechnologyTianjin300384ChinaAbstractSoundsourcelocalizationbasedonthetimedifferenceofarrival(TDOA)consistsoftwoparts-timedelayestimationandlocationestimation.Thesetwopartswillcausecumulativeerrorsandaffectthefinalaccuracyoflocalization.Aimingatthisproblem,thispaperproposesanimprovedaccumulatedmulti-framecrosspowerspectrumphase(CSP)timedelayestimationalgorithm,andonthisbasis,itimprovesthesphericalinterpolation(SI)algorithmtoderivethesourcelocation.Inordertoverifytheeffectivenessofthealgorithm,athree-dimensionalsoundsourcelocalizationsystemisestablishedunderthepracticalconditions.Thisresultofexperimentsshowsthattheproposedmethodcaneffectivelyreducetheerrorsandimprovethepositioningaccuracy.KeywordssoundsourcelocalizationTDOACSPSI2014-09-30.1989—E-maillsyecheng@163.com.1964—E-mailysying126@126.com..[1-2].3[3][4][5-6]..[7-8]..[9].1673-095X201502-0050-05JOURNALOFTIANJINUNIVERSITYOFTECHNOLOGY31220154Vol.31No.2Apr.2015201541VoiceactivitydetectionVAD.300~4kHz4kHz300Hz..[10-11].22.1xin=αisn-τi+εi1xjn=αjsn-τj+εj212snαiαjεiεjsnεiεjxinxjnRijτ=E[xinxjn-τ]=αiαjE[sn-τisn-τj-τ]3Rijτ=αiαjRsτ-τi-τjττ4Rssn4τ=τi-τjτ.Rijτ=12π∞-∞乙Gijωejωτdω5Gijωxixj..Rijgτ=12π∞-∞乙ψijωGijωejωτdω66ψijkψijk1/Gijω.τMiMj.τij=argmaxRijτττ72.2CSPCSP.CSP.ψacc-mcspf=flenk=1Σ1Gijkfp+min[rijk2f]8flenrijfrij2f]=Gijkf2GiifGjjf9ρ[0.51].σρ=ρ0σσ0[ρ0ρ1/σ0σ1]σ0σ1+ρ1σ0≤σ≤σ1ρ1σσ11111111111110σ0σ1ρ0ρ1ρ0ρ1.121Fig.1Thetraditionalcross-powerspectrumfunction2Fig.2Theimprovedcross-powerspectrumfunction-2000200001.00.50.0-0.5-1.0-2000200001.00.50.0-0.5-1.051··312..7.33.1、334.3.TDOA..M0、M1Q4.M0MiQRi=‖ri‖Rs=‖rs‖.riMiM0xiyizirsM0Qxsyszs.QMiDi=‖ri-rs‖=xi-xs2+yi-ys2+zi-zs2姨11Di=di0+Rs12di0=Di-D0MiM0Q.12R2i-di02-2Rsdi0-2riTrs=013di0ε=R2i-dij2-2Rsdij-2riTrs14M1MNM0N14Nε=δ-2Rsd-2Prs15δ=R21-d102R22-d202…R2N-dN02000000000000000000000000000000d=d10d20…dN0000000000000000000000000000000P=x1y1z1x2y2z2………xNyNzN000000000000000000000000000000RsrsRsrs=12P*wδ-2Rsd15P*w=PTP-1PTrsrs=12P*wδ-2RsdRsRsrs=12P*wδ-2Rsdr赞s。3.2Rsr赞sr赞s.15ε=Aθ-b16A=[P|d]θ=rsRs00b=12δ.143Fig.3ThegeometryofmicrophonearrayyM2M0M1θr准OM3zxMxyz4M0、M1QFig.4ThegeometricdiagramofM0M1andQM0MiRiQRs+di0Rs52··20154θ赞=ATA-1ATb17PTPPTddTPdTddddddddddddddddddd=QvvTkdd18v=-PTP-PTddTPdTd-1PTddTdQ=PTP-1[I-PTdvT]k=1-dTPvdTdPdPd=I-ddTdTd1918r赞sSI=PTPbP-1PTPdb20.4.USB2831450kHz.40000.5m0000.5m0000.5m..5.2.23.21.1η=r-r赞rrr赞.10%.....75°2~10m1m9252256.│准i-准│≤5°│ri-r│≤20cm25.准iri准r5Fig.5TheresultofVADstartend500015000100001.00.50.0-0.5-1.01Tab.1Experimentalresults/m/°/°/m/°/°/m/°/°244.6130.112.4347.2738.242.1845.7434.57468.2360.364.2770.2564.754.0868.9462.36675.6990.256.3776.4895.276.1476.4393.14879.28120.248.2479.94124.378.0979.77122.751081.50150.4710.4183.21156.1910.1782.26153.476Fig.6Therelationshipofrecognitionrateanddistance/m0.90.80.70.60.50.410684253··312.5~8m75%5m8m..0°180°15°6m1320260.7.76m0°~30°150°~180°60%.6m30°~150°70%85%..5..[1],,,.[J].,2009(3):199-205.[2],.[J].,2004(5):19-23.[3]FrancoisG,DominicL,FrancoisF,etal.TheManyEarsopenframework:microphonearrayopensoftwareandopenhardwaresystemforroboticapplications[J].AutonomousRobots,2013,34(3):217-232.[4]LarisaD,GuillermoP,IsmaelLL,etal.Materialsoundsourcelocalizationthroughheadphonesl[J].AcousticalPh-ysics,2012,58(5):610-617.[5],.[J].,2007(1):67-70.[6]XinwangWan,ZhenyangWu.Soundsourcelocalizationbasedondiscriminationofcross-correlationfunctions[J].AppliedAcoustics,2013,35(2):741-748.[7],,,.[J].,2013(8):1015-1019.[8]MehdiBD,HamidRA,MohammadRT.Soundsourcelo-calizationusingcompressivesensing-basedfeatureextrac-tionandspatialsparsity[J].DigitalSignalProcessing,2013,27(5):234-241.[9]RobertoC,AlessandraDeR.Sphericalinterpolationusingthepartitionofunitymethod:Anefficientandflexibleal-gorithm[J].AppliedMathematicsLetters,2012,63(2):2510-2519.[10].[J].,2006(1):142-144.[11]KentaroI,TomohiroN,MasakiyoF,etal.Noiserobustvoiceactivitydetectionbasedonperiodictoaperiodiccomponentratio[J].SpeechCommunication,2009,43(5):521-528.7Fig.7Therelationshipofrecognitionrateandazimuth/°1.00.80.60.40.205010015020054··
本文标题:基于到达时间差的声源定位算法与实现-叶诚
链接地址:https://www.777doc.com/doc-5149365 .html