《2、外文参考资料及译文-杨敏.docx》由会员分享，可在线阅读，更多相关《2、外文参考资料及译文-杨敏.docx（12页珍藏版）》请在三一办公上搜索。
1、金陂科技学院毕业设计（论文）外文参考资料及译文译文题目：基于语音识别的智能门控系统设计学生姓名：杨敏学号：0805110732专业：通信工程所在学院：信息技术学院指导老师：陈美君职称：高级工程师/副教授2011年12月7日说明：要求学生结合毕业设计（论文）课题参阅一篇以上的外文资料，并翻译至少一万印刷符（或译出3千汉字）以上的译文。译文原则上要求打印（如手写，一律用400字方格稿纸书写），连同学校供应的统一封面及英文原文装订，于毕业设计（论文）工作起先后2周内完成，作为成果考核的一部分。2009ISECSInternationalColloquiumonComputing.Communicat
2、ion.Control,andManagementDesignandRealizationofanIntelligentAccessControlSystemBasedonVoiceRecognitionBoCuiCollegeofInfbnnationHcbeiPolytechnicUniversityTangshan.ChinaE-mail:mikecuiTongzeXueCollegeofInfbnnationScienceandEngineeringHebeiUniversityofScienceandTechnologyShijiazhuang,ChinaE-inail:dianzi
3、xinxiAbstractTheintelligentaccesssystemadoptsthetechnologyofthevoicerecognitionthatbasedontheSPCE061Asinglechip.ThesystemhardwareismadeupofSPCE061Asinglechip,thepowerandgatingcircuit,theextendedmemorizerSPR4096,thevoiceinputandoutputcircuit.ThekeytechnologiesarctheapplicationoftheSPCE061Asinglechipo
4、nvoicerecognitionandthedesignofthedoorcontrolcircuit.Thesystemsoftwareismadeupofthevoicetrainingmodule,thevoicerecognitionmodule,thevoicedataprocessingmodule,thevoice-playingmoduleandthecodeofinputandoutputmodule.Thesystemcompletesthefunctionsofcollectingthevoicedata,distillingcharacter,specialvoice
5、recognitionandvoiceplayingintermsofinitializingthesystemandtheidcnticationtraining.Accordingtothevoicerecognitionarithmetictheory,thepretrcatmentofvoicesignal,thecharacterdistillingandpatternmatchingisanalyzed.Theresultsoftheexperimentsindicatethatthesystemcapabilityissteadyandtheidenticationiseffec
6、tive.ThesystemcanbeappliedinhouseorsmallofceSafetyprotection.Keywords-accesscontrol;linearprediction;voicerecognition;patternmatch;characterdistilI. Introduc11onTherearemanyidentificationtechnologiesusedincurrentintelligentguardsystem.Relativetoothertechniques,theVoiCerecognitiontechnologyisgenerall
7、yregardedasoneoftheconvenientandsaferecognitiontechniques;thistechniqueisakindoftechniquethatmakesuseofthecreaturecharacterofahumanbodytoCarryonidentification.Becauseeverybody,screaturecharacterisuniqueandstableinacertainperiod,theyarcdiIlcrcntfromothersanddifficulttobefabricatedandimitated.Sothevoi
8、cerecognitionteclmologycanbemadeuseofidentification,whichissafe,accurateanddependable.II. THEC1.ASSIFICATIONOFVOICERECOGNITIONBecausethepurposeandfunctionofthevoicerecognitionarcdiflcrcnt,therecognitionisclassifiedastalkerrecognitionandvoicerecognition.Andthetalkerrecognitioncanbeclassifiedastwotype
9、s,oneisrelevanttotextandtheotherisirrelevanttotext.Thevoicerecognitionsystemthatisrelevanttotextneedsuserstopronounceaccordingtothestatedcontents,andtheneverybodysspeechmodelisbuiltupaccurately.Becauseidentificationalsoneeduserstopronounceaccordingtothestatedcontents,theeffectisVerygood.Thevoicereco
10、gnitionsystemthatisirrelevanttotextdoesntrulepronunciationcontentsofthetalkers;itisditTiculttobuildupspeechmodels.Butcustomersmakeuseofthesystemconvenientlyanditisappliedwidely.Fromtheusage,thesystemcanbeclassifiedastalkerrecognitionandtalkerconfinnation.Theformerjudgesavoicethatneedstobeidentifiedf
11、romseveraltalkers.Thelatterjudgesthatanidcnticdvoicecomesfromacertaintalkerwhetherornot.Itsoutputonlyhashvokindsofresult,yesornot.ThecentralprocessorofthissystemistheSPCE061Asinglechip.Thetalkerconfirmationthatisrelevanttotextisrealizedonthechip,andthenhomologousorderandoperationarecarriedout.Thesys
12、temismainlymadeupoftalkeridentificationmodule,gatingcircuitanddoorlocketcIntraining,thevoiceoftalkersgetsintothevoicesignalcollectioncircuittIiroughamicrophone,andthenthecollectedvoicesignalsareprocessedbythevoiceprocessingcircuit,thecharacteristicparametersoftalkersaredistilledandsaved.Atlastthedat
13、abaseofcharacteristicparametersoftalkersisfb11ned.Inidentifying,thevoicethatneedstobeidentifiedismatchedwiththeinformationinthedatabaseofcharacteristicparametersoftalkers.Outputcircuitscontrolthegatingelectricalmachine,andlastlythedoorlockiscontrolled.Figurel.FrameofthissystemCCCM2009978-1-4244-4246
14、-109$25.002009IEEEThehardwarepartofthissystemincludeSPCE061Asinglechip,theaud100utputcircuit,theaudioinputcircuit,thevoicerecognitioncircuit,thedrlockcontrolcircuit,theF1.ASHcircuitandthekeyboardcircuitetc.TheframeofthissystemisshowedinFigure1.Themainmissionofthehardwareistochangevoicesignalsintodig
15、italsignals,collect,upload,save,identityandplaytheVOiCedatum.III. HARDWAREDESIGNOFTHESYSTEMA. CPUcorecircuitThevoltagescopeofaSPCE061Asinglechipis2.6-5.5V.thesystemfrequencyis0.375-49.152MHz.Tlierearemulti-functionprogrammableIOportsof32bits,tvvotimercountersof16bits.7channelsVoItageADCof10bits,micr
16、ophoneinputwithautomaticplusfunction,doublechannelsaud100utputfunctionof10bitsDACandWatchDogTimer(WDT)inthesinglechip.TheinterruptioncontrollercanhandletlreekindsofFIQinterruption,elevenkindsofIRQinterruptionandonesoftintemtioncontrolledbytheinstnctionBREAK.ITiereareabundantCRinctiondatabasesandvoic
17、eprocessingnctiondatabasesPrOVidedbytlesinglechip.Itisverysuitabletoenoldervoicerecognitionproducts.B. VoicerecognitioncircuitTheprincipleofvoicerecognitioncircuitsisthatoiccsignalsareanalyzedbytheintelligentsystemafterdistillingvoice,firstlynoisesarefilteredandtheusefulsignalsofvoicesignalsarcdisti
18、lledthroughafiltergroup,thenthesignalsareprocessedandchosenbycalculationmatchoffunctionPAR-COR.linearpredictioncoefficient,timesofzeroetc.ThevoicesignalscarryonmodematchwiththevoicedatuminthevoicedatabaseaHeranalysisandprocessing,lastlythevoicerecognitionresultisoutputaccordingtothematchresult.Theba
19、sicstructureofthevoicerecognitioncircuitisshowedinFigure2.Figre2ThevoicerecognitioncircuitC. PowersupplycircuitThechipSPCE061Aadoptslowvoltagesupplyinordertoreducepowerconsumption.TherearctvvopowersuppliesinSPCE061A,oneisinsidepowersupplyVDDandtheotheroneisI/OpowersupplyVDDH.I/Opowersupplyis5V.andin
20、sidepowersupplyis3.3V.Tolowertheinsidepowersupplyismainlytoreducepowerconsumptionandworkingtemperatureofthesinglechip.TlioughthevoltagerangeofSPCE061AisVerywide,tomakethechiprunmorestablyandsatisfythevoltagedemandof1.OportsandoutsidepartsthepowersupplycircuitisshowedinFigure3.AC220VisconcrtcdintoDC5
21、Vby7805steady-voltagemodule,DC5VwillsupplypowerforvoicerecognitionmoduleandeveryIOportsinsidesthesystem.DC5VisconvertedintoDC3.3VbyTR1972-33forCPUcore.IV. PROCESSINGOFTHEVOICESIGNA1.SThegeneralStruchireofprocessingofthevoicesignalsisshowedinFigure4.Figiirel.TliestnchreofvoicesignalsprocessingFigure4
22、showsus:firstlyvoicesignalsarepretreatedandthesignalsaeamplifiedproperly,secondlyanalogsignalsareconvertedintodigitalsignalsinorderthatdigitalequipmentscanprocessthesignalsConVenienUy,andthencharacteristicsaredistillediorderthatsomeparametersofthesignalcharacteristicscanbeusedtoreplacethevoicesignal
23、,lastlydifterenttreatmentwillbeadoptedaccordingthemissions.Voicerecognitioncanbedividedintohvostages:thetrainingstageandtheidentificationstage.Inthetrainingstagethevoicesignalsexpressedbycharacteristicparametersareprocessed,standarddatumthatcanshowcommoncharacteristicsofidentificationbasicunitsareob
24、tained.Referencetemplatesareformedbasedonabovedatum,andthenthereferencetemplatedatabaseisfo11nedafterallreferencetemplatesofidentifiedbasicunitsarecombinedtogether.IntheidentificationstagetheidentifiedVoiCesignalsaftercharacteristicdistillingarecomparedwithreferencetemplatesonebyone,andthenthevoicec
25、orrespondingtomostalikereferencetemplateisfoundout,thevoiceistheidentificationresult.A. VoicesignalspretreatmentThenoisesseriouslydisturbtheprocessingandidentificationofvoicesignals,sothenoisesmustbedisposedfirstly.Theinputanalogvoicesignalsfrommicrophonesmustbesampledandmeasuredinordertoobtaindigit
26、alvoicesignals.Beforeconertingvoicesignalsintodigitalsignals,itisnecessarytofilterandcounterdisturb.Infilteringthesignalpartandnoisesbeyond1/2samplefrequencyarcfiltered.Tliecleanvoicesignalsareobtainedlater,andthenIow-frcqucncydisturbingisfilteredthroughfbre-aggraxationtechnology,especiallythedistur
27、bingof50Hzor60Hz.Thehigl-frcqucncyvoicesignalsarcimproxedandtheycanremoveDCfloating,restrainrandomnoisesandimproethefunctionofenergyofcleanvoice.B. CharacteristicdistillingTliesystemadoptstheevaluationmethodthatusesthecontrastvaluebetweendispersiondegreeofdifferentspeakersandselfdispersiondegreeofea
28、chspeakerascharacteristicparameters.Tliebasicideais:todistillgroupcharacteristicparametersfromavoicesegmentofthesamespeaker,thatistosaytomapthesegmentonadotofthemulti-spacc.Diftcrcnlvoicesfromthesamespeakerwillproducedifterentdotsinthecharacteristicspace;thefunctionofmulti-variableprobabilitydensity
29、candescribethedistribution.Fordifferentsinglepronunciationfromthesamespeaker,thesedotsarerelativelyconcentrated.Butthepronunciationdistributionfiomdifferentspeakersisapartfarther,thegroupcharacteristicparameterscandescribethethumbprintofspeakerseffectively.Accordingtothisprinciple,forsingleparameter
30、,theFratiobetweentwokindsofdistributionparameterscanbeusedaseilectivemeasurementrule.TlieFratioshowsthecontrastbetweendispersiondegreeofdifterentspeakersandselfdispersiondegreeofeachspeaker.TheFraHoOfOnecharacteristicparameterisbigger,forthischaracteristic,thefb11nerisbiggerthanthelatteraveragely.Th
31、ereforetherecognitionsystemadoptsabiggerFratioandthenthesystemcapabilityisimproved.C. ModulematchAtpresentresearchonthemethodofmodulematchbasedonVarioUScharacteristicparametersismoreandmoreembedded.Typicalmethodsare:thevectormeasurearilhmeMc,theGaussmixturemodulearithmetic,thedynamictimewhole(DTM)ar
32、ithmeticandthemanualnervenetarithmetic.Theabovemethodshavebothadvantageandweakness.WhentheDTMarithmeticisappliedintheidentificationoflongvoice,theoperationsofmodulematcharetoogreat.ButthearithmeticissimpleandeffectiveforshortVoiCe(thelengthofvalidvoiceissubter-3seconds)identitcation.Sothemethodisesp
33、eciallyapplicabletothespeakerrecognitionsystemofshortvoiceandtext.ThesysteminthispaperadoptsIhCDTWarithmetic.V. EXPERIMENTRESU1.TSTheusersmustconfinnthepracticabilityandstabilityofthissystem.Thenhvoimportantparametersofhesystemcapabilityaretheerroridentificationratioandtherejectionidentificationrati
34、o.TheformeristhemistakemadebytheVoiCesignalsofnon-usersacceptedbythesystem,thelatteristhemistakemadebythevoicesignalsofusersrejectedbythesystem,theyarereleanttomatchtlresholdvalue.Ihesettingofthematchthresholdvalueisrelevanttothefunctionandapplicationofthevoicelocksystem.Forfamilyuserstheerroridenti
35、ficationratioandtherejectionidentificationrat100fthissystemmustbelikelylow,eveniszero.Firstlyacertainuser,svoiceisrecordedbyarecorder(thetone,thespeedandthecontentareallsame),thentherecordiscomparedwiththerealVoiCefortytimeswithindifterentdistance.ItisshowedinTable1.Afterfortyexperimentsbyrecorderth
36、etimeofpassageofuncertainusersiszerointheresults.TheexperimentresultisveryidealR)rfamilyusers.WhCnspeakersarcapartfromthemicrophonefarthertheidentificationratiodescendsobviously.Thereasonisthattherat100ftheenergyofusefulvoicesignalsandnoisesignalsdescendsgradually;distortionappearswhencharacteristic
37、distillingandlastlyleadstoinaccessibilityrequestsinmodulematch.Byadjustingthethresholdaluetheproblemcanbesolved.VI. CONC1.USIONSTheintelligentlocksysteminthispaperadoptsthevoicerecognitionteclnology.namelytheoicecontrolintelligentsystemoperatesthelocksystemandgivescorrespondingvoiceanswersandsuggest
38、ionsaccordingtothevoicesignals.Comparewithotherbiologyidentificationtechnologies,thevoiceidentificationteclmologyhassuchadvantagesasnon-loss.non-forgetting,convenienceetc.Andalsoitisacceptedbymanyusers,thecostofvoiceinputequipmentsislow.Thesystemcanbespreadandappliedconxcnicntlybecausethesystemdoesn
39、tinvolvetheusersprivacy.Experimenttestsshowthatthesystemisstableandeffective.TheinnoationinthepaperistoapplyvoiceprocessingteclmologyoftheSPCE061Asinglechipinthelocksystem,developtheusagescopeofsinglechips,enrichapplicationfieldofvoicerecognitionandprovideanewmethodtortheintelligentlocksystem.Refere
40、nces1 JiqingHail.1.ciZhang.TicranZlicng.VroiccSignalsProccssingM.Beijing:TsinglnuUniversityPress.20042 TicchcngYu.TlieDcvclopinenlStateOfthCVOiCCIdentificationJ.TheDevelopmentCoiuuiiUMcatiouWorld.2005.2:56-593 ShiqiangZhao.BingliJin.ManZhao.ApplicatiouandExploitationofthe16BitsSingleChipProcessorSPC
41、E06IAJ.IiitemationalElectiouicElements.2003.5:37-394 ZhilinsJiang.TlieSystemofDataMiiltiprogniinnuugCollectionandTrausiuissionUsiug1.iugyaugSiuglechipComputeiJ.SichuauUniversityofScienceandTeclmology.200524(2):69-725 JimgiiAng1.uo,RongjiangTaug.ResearchaudRealizationofSpeaker-IndependentSpeeChRecogn
42、itionSystembasedonDSPJ.ControlandAutomationPublicationGroup.2007.4-2:195-1966 MohamedAbouiiana.OnlineweightandshrinkagecontrolofcottonknitsEBO1.).btt,www.ntcresearch.orgdfrpUAnRp03S01-PH07.A3.pdf.24-12-107 XiaoningDong.MengZhaug.DesignofVoiCeRecognitionSystemBasedon1.ingyangSiugleChipJ.ElectrouicEng
43、ineer.2008.2:18-8 DanhiiaChang.ClHinlCiZheng.ControlliugofbaiupowerofSPeeerecognitionsystembasedonDSPJ.ElectronicMeasuremeTecluiology.2008.4:175-1789 TaoHan,1.ingWang.Hui1.iu.AnImprovedSpeechEndpoiDetectionMethodBasedonSpectralEntropyJ.MicioelectrouicsComputer.2008.5:146-14910 FengYmg.YuienFeng.Desi
44、gnofSpeechRecognifiouSystemBa：onSPCE061AJ.MiciocoiiiputerIufbniuitioii.2008.6-2:121“23Figure3.PowersupplycircuitTAB1.ElTHEMATCHRESU1.TOFRECORDANDREA1.VOICEFROMACERTAINUSERVoiceclasTimesThedistanceofuserandmicrophoneThesamevoiceofacertainuserTherecordvoiceofacertainuserThealikevoiceofacertainuserThes
45、amevoiceofanuncertainuserThealikevoiceofanuncertainuserRejectiontimesIm3740364039Acceptancetimesln3O4O1Rejectiontimes2n2537394040Acceptancetimes22m1531OO基于语音识别的智能门控系统设计崔波中国，唐山河北理工高校信息学院邮箱：mikecuiheut.edu薛同泽中国，石家庄河北科技高校信息科学与工程学院邮箱：dianzixinxi摘要：该智能门控系统采纳语音识别技术，在陵阳单片机SPCEo61A上实现。硬件部分主要由SPCEO61A单片机、电源与
46、门控电路、外扩存储器SPR4096、音频输入与输出电路等组成。主要技术是该单片机在语音识别中的应用和对门控电路的设计。该系统软件部分由语音训练模块、语音识别模块、语音数据处理模块、语音播报模块和密码输入与确认模块组成，通过对系统的初始化和识别训练实现了语音数据的采集、特征提取、特定语音识别和语音播放等功能。依据语音识别算法原理，对语音信号的预处理、特征提取、模式匹配做了分析。该系统试验结果表明，系统性能稳定，识别效果好，可用于家庭或小型办公场所的门锁限制。关键词：访问限制、线性预料、语音识别、模式匹配、特征提取一、引言当前智能型门卫系统中利用的识别技术有许多种，其中基于特定人的身份识别的技术也
47、有许多，但相对于其他技术，语音识别技术是目前公认的比较便利与平安的识别技术之一，该技术是利用人体生物特征进行身份认证的一种技术。由于每个人的生物特征与其他人相比具有不同的唯一性和在确定时期内不变的稳定性，因此不易伪造和假冒，所以利用语音识别技术进行身份认证，平安、精确、牢靠。二、语音识别的分类由于语音识别的目的和作用不同，识别又分为说话人识别和语音识别。其中说话人识别，可分为与文本有关和与文本无关两类。与文本有关的语音识别系统要求用户依据规定的内容发音，每个人的语音模型逐个被精确地建立，而识别时也必需按规定的内容发音，因此可以达到较好的识别效果；而与文本无关的识别系统则不规定说话人的发音内容，
48、模型建立相对困难，但用户运用便利，应用范围较宽。从用途上看，可分为说话人识别和说话人确认。前者判定某一待识别的声音是多个话者中的一个，是多选一的问题，属于闭集识别范围。后者判定一个待识别的声音“是或不是”某一特定说话者的语音，其输出只有两种结果，为确定或否定的问题。该系统的核心处理器是陵阳16位单片机SPCEO61A,并在其上实现与文本有关的说话人的确认，然后执行相应的吩咐与操作。该系统主要由说话人识别模块、门锁限制电机以及门锁等部分组成。在训练时，说话人的声音通过麦克风进入说话人语音信号采集前端电路,由语音信号处理电路对采集的语音信号进行特征化和语音处理，提取说话人的特性特征参数并进行存储，形成说话人特征参数数据库。在识别时，将待识别语音与说话人特征参数数据库进行匹配，通过输出电路限制门控电机，最终实现对门锁的限制。图1系统结构框图本系统的硬件部分由单片机SPCEO61A、音频输出电路、音频输入电路、语音识别电路
