搜索
DigitalSignalProcessing数字信号处理院系:专业:学号:姓名:【英文原文】DigitalSignalProcessing1、IntroductionDigitalsignalprocessingiswillsignaltodigitallysaysanddealwiththetheoryandtechnology.Digitalsignalprocessingandanalogsignalprocessingissignalprocessingsubset.Digitalsignalprocessingalgorithmneedtousespecialprocessingequipmentsuchascomputerordigitalsignalprocessorandapplication-specificintegratedcircuits,etc.Digitalsignalprocessingtechnologyandequipmentwithflexible,preciesanti-jammingofstrong,equipmentofsmallsize,lowcost,speedsuchoutstandingadvantages,thesearesimulationsignalprocessingtechnologyandequipmentandincomparable.SincethegoalofDSPisusuallytomeasureorfiltercontinuousreal-worldanalogsignals,thefirststepisusuallytoconvertthesignalfromananalogtoadigitalform,byusingananalogtodigitalconverter.Often,therequiredoutputsignalisanotheranalogoutputsignal,whichrequiresadigitaltoanalogconverter.Evenifthisprocessismorecomplexthananalogprocessingandhasadiscretevaluerange,thestabilityofdigitalsignalprocessingthankstoerrordetectionandcorrectionandbeinglessvulnerabletonoisemakesitadvantageousoveranalogsignalprocessingformany,thoughnotall,applications.DSPalgorithmshavelongbeenrunonstandardcomputers,onspecializedprocessorscalleddigitalsignalprocessors(DSP)s,oronpurpose-builthardwaresuchasapplication-specificintegratedcircuit(ASICs).Todaythereareadditionaltechnologiesusedfordigitalsignalprocessingincludingmorepowerfulgeneralpurposemicroprocessors,field-programmablegatearrays(FPGAs),digitalsignalcontrollers(mostlyforindustrialapplicationssuchasmotorcontrol),andstreamprocessors,amongothers.InDSP,engineersusuallystudydigitalsignalsinoneofthefollowingdomains:timedomain(one-dimensionalsignals),spatialdomain(multidimensionalsignals),frequencydomain,autocorrelationdomain,andwaveletdomains.Theychoosethedomaininwhichtoprocessasignalbymakinganinformedguess(orbytryingdifferentpossibilities)astowhichdomainbestrepresentstheessentialcharacteristicsofthesignal.Asequenceofsamplesfromameasuringdeviceproducesatimeorspatialdomainrepresentation,whereasadiscreteFouriertransformproducesthefrequencydomaininformationthatisthefrequencyspectrum.Autocorrelationisdefinedasthecross-correlationofthesignalwithitselfovervaryingintervalsoftimeorspace.2、SignalSamplingWiththeincreasinguseofcomputerstheusageofandneedfordigitalsignalprocessinghasincreased.Inordertouseananalogsignalonacomputeritmustbedigitizedwithananalogtodigitalconverter(ADC).Samplingisusuallycarriedoutintwostages,discretizationandquantization.Inthediscretizationstage,thespaceofsignalsispartitionedintoequivalenceclassesandquantizationiscarriedoutbyreplacethesignalwithrepresentativesignalvaluesareapproximatedbyvaluesfromafiniteset.TheNyquist-Shannonsamplingtheoremstatesthatasignalcanbeexactlyreconstructedfromitssamplesifthesamplesifthesamplingfrequencyisgreaterthantwicethehighestfrequencyofthesignal.Inpractice,thesamplingfrequencyisoftensignificantlymorethantwicetherequiredbandwidth.Adigitaltoanalogconverter(DAC)isusedtoconvertthedigitalsignalbacktoanalogsignal.Theuseofadigitalcomputerisakeyingredientindigitalcontrolsystems.3、TimeandSpaceDomainsThemostcommonprocessingapproachinthetimeorspacedomainisenhancementoftheinputsignalthroughamethodcalledfiltering.Filteringgenerallyconsistsofsometransformationofanumberofsurroundingsamplesaroundthecurrentsampleoftheinputoroutputsignal.Therearevariouswaystocharacterizefilters,forexample:A“linear”filterisalineartransformationofinputsamples;otherfiltersare“non-linear.”Linearfilterssatisfythesuperpositioncondition,i.e.ifaninputisaweightedlinearcombinationofdifferentsignals,theoutputisanequallyweightedlinearcombinationofthecorrespondingoutputsignals.A“causal”filterusesonlyprevioussamplesoftheinputoroutputsignals;whilea“non-causal”filterusesfutureinputsamples.Anon-causalfiltercanusuallybechangedintoacausalfilterbyaddingadelaytoit.A“time-invariant”filterhasconstantpropertiesovertime;otherfilterssuchasadaptivefilterschangeintime.Somefiltersare“stable”,othersare“unstable”.Astablefilterproducesanoutputthatconvergestoaconstantvaluewithtime,orremainsboundedwithinafiniteinterval.Anconvergestoaconstantvaluewithtime,orremainsboundedwithinafiniteinterval.Anunstablefiltercanproduceanoutputthatgrowswithoutbounds,withboundedorevenzeroinput.A“FiniteImpulseResponse”(FIR)filterusesonlytheinputsignal,whilean“InfiniteImpulseResponse”filter(IIR)usesboththeinputsignalandprevioussamplesoftheoutputsignal.FIRfiltersarealwaysstable,whileIIRfiltersmaybeunstable.MostfilterscanbedescribedinZ-domain(asupersetofthefrequencydomain)bytheirtransferfunctions.Afiltermayalsobedescribedasadifferenceequation,acollectionofzeroesandpolesor,ifitisanFIRfilter,animpulseresponseorstepresponse.TheoutputofanFIRfiltertoanygiveninputmaybecalculatedbyconvolvingtheinputsignalwiththeimpulseresponse.Filterscanalsoberepresentedbyblockdiagramswhichcanthenbeusedtoderiveasampleprocessingalgorithmtoimplementthefilterusinghardwareinstructions.4、FrequencyDomainSignalsareconvertedfromtimeorspacedomaintothefrequencydomainusuallythroug