SIST EN 300 961 V8.0.2:2003
(Main)Digital cellular telecommunications system (Phase 2+) (GSM); Full rate speech; Transcoding (GSM 06.10 version 8.0.2 Release 1999)
- Abstract
Updated to version 8.0.0 CRs from SMG#31
- Status
- Published
- Publication Date
- 30-Nov-2003
- Current Stage
- 6060 - National Implementation/Publication (Adopted Project)
- Start Date
- 01-Dec-2003
- Due Date
- 01-Dec-2003
- Completion Date
- 01-Dec-2003
SIST EN 300 961 V8.0.2:2003 is the Slovenian adoption of ETSI EN 300 961 V8.0.2 for GSM full rate speech transcoding. It specifies the bit-exact speech codec used in the GSM Phase 2+ digital cellular system, including the PCM interface, the encoder and decoder algorithm, and the test sequences used to verify implementations. The document is written for equipment designers, test engineers, and conformity teams working on GSM full-rate speech paths.
What does SIST EN 300 961 V8.0.2:2003 specify?
SIST EN 300 961 V8.0.2:2003 specifies the full rate speech transcoding function for the GSM Traffic Channel (TCH). It maps blocks of 160 speech samples in 13-bit uniform PCM format to 260-bit encoded frames and back again, at 8 kHz and an average encoded bit rate of 13 kbit/s.
The document also covers conversion between 8-bit A-law or µ-law (PCS1900) PCM and 13-bit uniform PCM, but only as part of the speech transcoding chain. It is organized into six clauses and three annexes.
| Annex | What it covers |
|---|---|
| Annex A | Codec performance, speech and non-speech behavior, delay, subjective bit relevance, and test sequence file format |
| Annex B | Test sequence disks |
| Annex C | Change Request History |
What are the key requirements of SIST EN 300 961 V8.0.2:2003?
SIST EN 300 961 V8.0.2:2003 defines a bit-exact codec chain that implementers must follow for interoperable GSM full rate speech handling. In practice, this means the encoder, decoder, fixed-point arithmetic, and test behavior all have to match the specified procedures.
Frame structure and signal path
Clause 1 defines a 20 ms speech frame of 160 samples, split into four 40-sample sub-segments. The encoder produces 260 bits per frame, and the decoder reconstructs 160 samples per frame. This matters because timing, buffering, and frame alignment in the equipment must match GSM speech processing.
Audio interface and PCM conversion
Clause 1 describes the analogue-to-digital and digital-to-analogue path, including input filtering, 8 kHz sampling, 13-bit two's complement uniform PCM, reconstruction filtering, and output level adjustment. It also defines A-law and µ-law conversion using the stated PCM procedures. This is the practical interface rule for terminal equipment and network-side conversion points.
Speech coding method
Clause 3 describes the Regular Pulse Excitation-Long Term prediction-Linear Predictive Coder (RPE-LTP). The encoder applies offset compensation, pre-emphasis, LPC analysis, Long Term Prediction (LTP), RPE grid selection, and Adaptive Pulse Code Modulation (APCM). The decoder performs the inverse operations plus de-emphasis. For implementers, this is the core algorithm that must be reproduced exactly.
Parameter coding and priority
Clause 1.7 defines the order and allocation of the encoded bits for the 260-bit speech frame, including Log.-Area Ratios (LAR), LTP lag and gain, grid position, block amplitude, and RPE pulses. The bit ordering also reflects subjective importance before channel coding in GSM 05.03. That matters when protecting bits with unequal importance over the radio path.
Delay and transmission behavior
Clause 2 states that the transcoder delay should be less than 30 ms, with a theoretical minimum of 20 ms. It also points to out-of-band signal discrimination and spurious out-of-band signal requirements for the analogue/digital interfaces. This is the design target for real-time voice equipment.
Fixed-point implementation
Clause 5 specifies the arithmetic model: 16-bit integers, 32-bit long integers, saturation, rounding, scaling, and named operations such as add, sub, mult, and norm. This lets independent implementations produce the same bit-exact results.
Verification and homing
Clause 4 defines code homing, encoder-homing frames, decoder-homing frames, and the required internal homestate behavior. Clause 6 defines digital test sequences for encoder-only, decoder-only, homing, bit synchronization, and frame synchronization testing. In practice, this is what laboratories use to prove conformance.
What terms does SIST EN 300 961 V8.0.2:2003 define?
- RPE-LTP - Regular Pulse Excitation-Long Term prediction-Linear Predictive Coder, the GSM full rate speech coding method used in this document.
- LTP - Long Term Prediction, the stage that estimates periodic speech structure over each 40-sample sub-segment.
- LAR - Log.-Area Ratios, the coded form of the LPC reflection coefficients.
- APCM - Adaptive Pulse Code Modulation, used to code the selected RPE sequence within each sub-segment.
- TCH - Traffic Channel, the GSM channel type for which the transcoding procedure applies.
- PCM - Pulse Code Modulation, the 13-bit uniform and 8-bit A-law or µ-law sample formats used at the interfaces.
- code homing - A test feature that resets encoder or decoder state by sending specific homing frames.
Who uses SIST EN 300 961 V8.0.2:2003?
SIST EN 300 961 V8.0.2:2003 is used by GSM speech codec developers, mobile station and network equipment manufacturers, and test laboratories. It is also used by conformity and type-approval teams that need bit-exact encoder and decoder verification.
In practice, users apply it to implement the GSM full rate speech codec, check fixed-point behavior, run the supplied digital test sequences, and verify homing and synchronization behavior. It is also relevant where TRAU testing or A-law and µ-law PCM interfaces are involved.
What changed in SIST EN 300 961 V8.0.2:2003 from the previous edition?
The scope notes that this version was updated to version 8.0.0 CRs from SMG#31. The change history also records the introduction of homing and 8-bit test sequences, the addition of µ-law (PCS1900), inclusion of associated ZIP files, and the publication update to version 8.0.2.
Which standards are used with SIST EN 300 961 V8.0.2:2003?
- GSM01.04 - abbreviations and acronyms used in the document.
- GSM05.03 - channel coding, including the bit rearrangement after speech coding.
- GSM06.01 - the reference speech transmission chain and the context for the transcoder.
- GSM11.10 - conformity specification for mobile stations, including performance aspects.
- ETS 300 085 - handset attachment requirements used in the interface context.
- ITU-T G.711 - A-law and PCM telephony conversion reference.
- ITU-T G.712 - transmission performance characteristics for PCM systems.
- ITU-T G.726 - ADPCM conversion procedures used for A-law and µ-law handling.
- ITU-T Q.35 - technical characteristics of telephony tones.
- ITU-T V.21 and ITU-T V.23 - modem-related tone/data references.
- GSM06.32 - Voice Activity Detector (VAD), referenced for homing state variables.
What does the SIST EN 300 961 V8.0.2:2003 document contain?
The document contains the encoder bit allocation table, block diagrams of the encoder and decoder, detailed functional descriptions, and fixed-point arithmetic procedures for every major processing step. It also includes the coded parameter tables, gain and quantization tables, homing-frame definitions, and test sequence formats.
Annex A adds performance discussion, including speech and non-speech behavior, delay, and the subjective relevance of output bits. Annex B identifies the supplied test sequence disks, and Annex C records the change request history.
Get Certified
Connect with accredited certification bodies for this standard
ANCE
Mexican certification and testing association.
Intertek Slovenia
Intertek testing, inspection, and certification services in Slovenia.
LNE (Laboratoire National de Métrologie et d'Essais)
French national laboratory for metrology and testing.
Sponsored listings
Frequently Asked Questions
SIST EN 300 961 V8.0.2:2003 is a standard published by the Slovenian Institute for Standardization (SIST). Its full title is "Digital cellular telecommunications system (Phase 2+) (GSM); Full rate speech; Transcoding (GSM 06.10 version 8.0.2 Release 1999)". This standard covers: Updated to version 8.0.0 CRs from SMG#31
Updated to version 8.0.0 CRs from SMG#31
SIST EN 300 961 V8.0.2:2003 is classified under the following ICS (International Classification for Standards) categories: 33.070.50 - Global System for Mobile Communication (GSM). The ICS classification helps identify the subject area and facilitates finding related standards.
SIST EN 300 961 V8.0.2:2003 is available in PDF format for immediate download after purchase. The document can be added to your cart and obtained through the secure checkout process. Digital delivery ensures instant access to the complete standard document.
Standards Content (Sample)
2003-01.Slovenski inštitut za standardizacijo. Razmnoževanje celote ali delov tega standarda ni dovoljeno.Digital cellular telecommunications system (Phase 2+) (GSM); Full rate speech; Transcoding (GSM 06.10 version 8.0.2 Release 1999)33.070.50Globalni sistem za mobilno telekomunikacijo (GSM)Global System for Mobile Communication (GSM)ICS:Ta slovenski standard je istoveten z:EN 300 961 Version 8.0.2SIST EN 300 961 V8.0.2:2003en01-december-2003SIST EN 300 961 V8.0.2:2003SLOVENSKI
STANDARD
ETSIEN300961V8.0.2(2000-11)EuropeanStandard(Telecommunicationsseries)Digitalcellulartelecommunicationssystem(Phase2+);Fullratespeech;Transcoding(GSM06.10version8.0.2Release1999)GLOBALSYSTEMFORMOBILECOMMUNICATIONSRSIST EN 300 961 V8.0.2:2003
ETSIETSIEN300961V8.0.2(2000-11)2(GSM06.10version8.0.2Release1999)ReferenceREN/SMG-110610Q8KeywordsDigitalcellulartelecommunicationssystem,GlobalSystemforMobilecommunications(GSM)ETSI650RoutedesLuciolesF-06921SophiaAntipolisCedex-FRANCETel.:+33492944200Fax:+33493654716SiretN°34862356200017-NAF742CAssociationàbutnonlucratifenregistréeàlaSous-PréfecturedeGrasse(06)N°7803/88ImportantnoticeIndividualcopiesofthepresentdocumentcanbedownloadedfrom:http://www.etsi.orgThepresentdocumentmaybemadeavailableinmorethanoneelectronicversionorinprint.Inanycaseofexistingorperceiveddifferenceincontentsbetweensuchversions,thereferenceversionisthePortableDocumentFormat(PDF).Incaseofdispute,thereferenceshallbetheprintingonETSIprintersofthePDFversionkeptonaspecificnetworkdrivewithinETSISecretariat.Usersofthepresentdocumentshouldbeawarethatthedocumentmaybesubjecttorevisionorchangeofstatus.InformationonthecurrentstatusofthisandotherETSIdocumentsisavailableathttp://www.etsi.org/tb/status/Ifyoufinderrorsinthepresentdocument,sendyourcommentto:editor@etsi.frCopyrightNotificationNopartmaybereproducedexceptasauthorizedbywrittenpermission.Thecopyrightandtheforegoingrestrictionextendtoreproductioninallmedia.©EuropeanTelecommunicationsStandardsInstitute2000.Allrightsreserved.SIST EN 300 961 V8.0.2:2003
ETSIETSIEN300961V8.0.2(2000-11)3(GSM06.10version8.0.2Release1999)ContentsIntellectualPropertyRights.6Foreword.61Scope.81.1References.81.1.1Abbreviations.91.2Outlinedescription.91.3Functionaldescriptionofaudioparts.91.4PCMFormatconversion.101.5PrinciplesoftheRPE-LTPencoder.101.6PrinciplesoftheRPE-LTPdecoder.111.7Sequenceandsubjectiveimportanceofencodedparameters.112Transmissioncharacteristics.142.1Performancecharacteristicsoftheanalogue/digitalinterfaces.142.2Transcoderdelay.143FunctionaldescriptionoftheRPE-LTPcodec.143.1FunctionaldescriptionoftheRPE-LTPencoder.143.1.1Offsetcompensation.153.1.2Pre-emphasis.153.1.3Segmentation.153.1.4Autocorrelation.163.1.5SchurRecursion.163.1.6TransformationofreflectioncoefficientstoLog.-AreaRatios.163.1.7QuantizationandcodingofLog.-AreaRatios.163.1.8DecodingofthequantizedLog.-AreaRatios.173.1.9InterpolationofLog.-AreaRatios.173.1.10TransformationofLog.-AreaRatiosintoreflectioncoefficients.173.1.11Shorttermanalysisfiltering.173.1.12Sub-segmentation.183.1.13CalculationoftheLTPparameters.183.1.14Coding/DecodingoftheLTPlags.183.1.15Coding/DecodingoftheLTPgains.193.1.16Longtermanalysisfiltering.193.1.17Longtermsynthesisfiltering.193.1.18WeightingFilter.203.1.19AdaptivesampleratedecimationbyRPEgridselection.203.1.20APCMquantizationoftheselectedRPEsequence.203.1.21APCMinversequantization.213.1.22RPEgridpositioning.223.2Decoder.223.2.1RPEdecodingsection.223.2.2LongTermPredictionsection.223.2.3Shorttermsynthesisfilteringsection.223.2.4Post-processing.224Codechoming.264.1Functionaldescription.264.2Definitions.264.3Encoderhoming.274.4Decoderhoming.274.5Encoderhomestate.284.6Decoderhomestate.285ComputationaldetailsoftheRPE-LTPcodec.285.1Datarepresentationandarithmeticoperations.285.2FixedpointimplementationoftheRPE-LTPcoder.30SIST EN 300 961 V8.0.2:2003
ETSIETSIEN300961V8.0.2(2000-11)4(GSM06.10version8.0.2Release1999)5.2.0Scalingoftheinputvariable.315.2.1Downscalingoftheinputsignal.315.2.2Offsetcompensation.315.2.3Pre-emphasis.315.2.4Autocorrelation.315.2.5Computationofthereflectioncoefficients.325.2.6TransformationofreflectioncoefficientstoLog.-AreaRatios.335.2.7QuantizationandcodingoftheLog.-AreaRatios.335.2.8DecodingofthecodedLog.-AreaRatios.345.2.9Computationofthequantizedreflectioncoefficients.345.2.9.1InterpolationoftheLARpp[1.8]togettheLARp[1.8].345.2.9.2Computationoftherp[1.8]fromtheinterpolatedLARp[1.8].355.2.10Shorttermanalysisfiltering.355.2.11CalculationoftheLTPparameters.355.2.12Longtermanalysisfiltering.375.2.13Weightingfilter.375.2.14RPEgridselection.385.2.15APCMquantizationoftheselectedRPEsequence.385.2.16APCMinversequantization.395.2.17RPEgridpositioning.395.2.18Updateofthereconstructedshorttermresidualsignaldp[-120.-1].405.3FixedpointimplementationoftheRPE-LTPdecoder.405.3.1RPEdecodingsection.405.3.2Longtermsynthesisfiltering.405.3.3Computationofthedecodedreflectioncoefficients.415.3.4Shorttermsynthesisfilteringsection.415.3.5De-emphasisfiltering.425.3.6Upscalingoftheoutputsignal.425.3.7Truncationoftheoutputvariable.425.4TablesusedinthefixedpointimplementationoftheRPE-LTPcoderanddecoder.436Digitaltestsequences.446.1Inputandoutputsignals.446.2Configurationfortheapplicationofthetestsequences.446.2.1Configuration1(encoderonly).456.2.2Configuration2(decoderonly).456.3Testsequences.466.3.1Testsequencesforconfiguration1.466.3.2Testsequencesforconfiguration2.466.3.3AdditionalTestsequencesforCodecHoming.506.3.3.1Codechomingframes.506.3.3.2Sequenceforanextensivetestofthedecoderhoming.506.3.3.3Sequencesforfindingthe20msframingoftheGSMfullratespeechencoder.506.3.3.4Formatsandsizesofthesynchronizationsequences.51AnnexA(informative):Codecperformance.53A.1PerformanceoftheRPE-LTP.53A.1.1Introduction.53A.1.2Speechperformance.53A.1.2.1Singleencoding.53A.1.2.2Speechperformancewheninterconnectedwithcodingsystemsonananaloguebasis.54A.1.2.2.1Performancewith32kbit/sADPCM(G.721,supersededbyG.726).54A.1.2.2.2PerformancewithanotherRPE-LTPcodec.54A.1.2.2.3PerformancewithencodingotherthanRPE-LTPand32kbit/sADPCM(G.721,supersededbyG.726).54A.1.3Non-speechperformance.55A.1.3.1Performancewithsinglesinewaves.55A.1.3.2PerformancewithDTMFtones.55A.1.3.3Performancewithinformationtones.55A.1.3.4Performancewithvoice-banddata.55A.1.4Delay.55A.1.5Bibliography.57SIST EN 300 961 V8.0.2:2003
ETSIETSIEN300961V8.0.2(2000-11)5(GSM06.10version8.0.2Release1999)A.2Subjectiverelevanceofthespeechcoderoutputbits.57A.3Formatfortestsequencedistribution.59A.3.1Typeoffilesprovided.59A.3.2Fileformatdescription.60AnnexB(informative):Testsequencedisks.62AnnexC(informative):ChangeRequestHistory.63History.64SIST EN 300 961 V8.0.2:2003
ETSIETSIEN300961V8.0.2(2000-11)6(GSM06.10version8.0.2Release1999)IntellectualPropertyRightsIPRsessentialorpotentiallyessentialtothepresentdocumentmayhavebeendeclaredtoETSI.TheinformationpertainingtotheseessentialIPRs,ifany,ispubliclyavailableforETSImembersandnon-members,andcanbefoundinETSISR000314:"IntellectualPropertyRights(IPRs);Essential,orpotentiallyEssential,IPRsnotifiedtoETSIinrespectofETSIstandards",whichisavailablefromtheETSISecretariat.LatestupdatesareavailableontheETSIWebserver(http://www.etsi.org/ipr).PursuanttotheETSIIPRPolicy,noinvestigation,includingIPRsearches,hasbeencarriedoutbyETSI.NoguaranteecanbegivenastotheexistenceofotherIPRsnotreferencedinETSISR000314(ortheupdatesontheETSIWebserver)whichare,ormaybe,ormaybecome,essentialtothepresentdocument.ForewordThisEuropeanStandard(Telecommunicationsseries)hasbeenproducedbyETSITechnicalCommitteeSpecialMobileGroup(SMG).Thepresentdocumentspecifiesthefullratespeechtranscodingwithinthedigitalcellulartelecommunicationssystem.NOTE:ThepresentdocumentisareproductionofrecommendationT/L/03/11"13kbit/sRegularPulseExcitation-LongTermPrediction-LinearPredictiveCoderforuseinthedigitalcellulartelecommunicationssystem".Archiveen_300961v080002p0.ZIPwhichaccompaniesthepresentdocument,containstestsequences,asdescribedinclause6andannexA.3.Thearchivecontainsthefollowing:Disk1.zipAnnexB:TestsequencesfortheGSMFullRatespeechcodec;TestsequencesSEQ01.xxxtoSEQ05.xxx.(Disk1.zipcontainsLHAcompressedfiles.)Disk2.zipAnnexB:TestsequencesfortheGSMFullRatespeechcodecwithhomingframes;TestsequencesSEQ01H.*toSEQ02H.*.Disk3.zipAnnexB:TestsequencesfortheGSMFullRatespeechcodecwithhomingframes;TestsequencesSEQ03H.*toSYNC159.COD.Disk4.zipAnnexB:8bitA-lawtestsequencesfortheGSMFullRatespeechcodecwithandwithouthomingframes(Disk4.zipcontainsself-extractingfiles).Disk5.zipAnnexB:8bitµ-lawtestsequencesfortheGSMFullRatespeechcodecwithandwithouthomingframes(Disk5.zipcontainsself-extractingfiles).SIST EN 300 961 V8.0.2:2003
ETSIETSIEN300961V8.0.2(2000-11)7(GSM06.10version8.0.2Release1999)ThecontentsofthepresentdocumentissubjecttocontinuingworkwithinSMGandmaychangefollowingformalSMGapproval.ShouldSMGmodifythecontentsofthepresentdocumentitwillbere-releasedwithanidentifyingchangeofreleasedateandanincreaseinversionnumberasfollows:Version8.x.ywhere:8indicatesRelease1999ofGSMPhase2+xtheseconddigitisincrementedforallchangesofsubstance,i.e.technicalenhancements,corrections,updates,etc.ythethirddigitisincrementedwheneditorialonlychangeshavebeenincorporatedinthespecification.NationaltranspositiondatesDateofadoptionofthisEN:3November2000DateoflatestannouncementofthisEN(doa):28February2001DateoflatestpublicationofnewNationalStandardorendorsementofthisEN(dop/e):31August2001DateofwithdrawalofanyconflictingNationalStandard(dow):31August2001SIST EN 300 961 V8.0.2:2003
ETSIETSIEN300961V8.0.2(2000-11)8(GSM06.10version8.0.2Release1999)1ScopeThetranscodingprocedurespecifiedinthepresentdocumentisapplicableforthefull-rateTrafficChannel(TCH)inthedigitalcellulartelecommunicationssystem.Theuseofthistranscodingschemeforotherapplicationshasnotbeenconsidered.InGSM06.01,areferenceconfigurationforthespeechtransmissionchainofthedigitalcellulartelecommunicationssystemisshown.Accordingtothisreferenceconfiguration,thespeechencodertakesitsinputasa13bituniformPCMsignaleitherfromtheaudiopartofthemobilestationoronthenetworkside,fromthePSTNviaan8bit/A-orµ-law(PCS1900)to13bituniformPCMconversion.TheencodedspeechattheoutputofthespeechencoderisdeliveredtoachannelencoderunitwhichisspecifiedinGSM05.03.Inthereceivedirection,theinverseoperationstakeplace.Thepresentdocumentdescribesthedetailedmappingbetweeninputblocksof160speechsamplesin13bituniformPCMformattoencodedblocksof260bitsandfromencodedblocksof260bitstooutputblocksof160reconstructedspeechsamples.Thesamplingrateis8000sample/sleadingtoanaveragebitratefortheencodedbitstreamof13kbit/s.Thecodingschemeistheso-calledRegularPulseExcitation-LongTermprediction-LinearPredictiveCoder,here-afterreferredtoasRPE-LTP.ThepresentdocumentalsospecifiestheconversionbetweenA-andµ-law(PCS1900)PCMand13bituniformPCM.Performancerequirementsfortheaudioinputandoutputpartsareincludedonlytotheextentthattheyaffectthetranscoderperformance.Thepresentdocumentalsodescribesthecodecdowntothebitlevel,thusenablingtheverificationofcompliancetothepresentdocumenttoahighdegreeofconfidencebyuseofasetofdigitaltestsequences.Thesetestsequencesaredescribedandarecontainedinarchiveen_300961v080002p0.ZIPwhichaccompaniesthepresentdocument.1.1ReferencesThefollowingdocumentscontainprovisionswhich,throughreferenceinthistext,constituteprovisionsofthepresentdocument.• Referencesareeitherspecific(identifiedbydateofpublication,editionnumber,versionnumber,etc.)ornon-specific.• Foraspecificreference,subsequentrevisionsdonotapply.• Foranon-specificreference,thelatestversionapplies.• Anon-specificreferencetoanETSshallalsobetakentorefertolaterversionspublishedasanENwiththesamenumber.• ForthisRelease1999document,referencestoGSMdocumentsareforRelease1999versions(version8.x.y).[1]GSM01.04:"Digitalcellulartelecommunicationssystem(Phase2+);Abbreviationsandacronyms".[2]GSM05.03:"Digitalcellulartelecommunicationssystem(Phase2+);Channelcoding".[3]GSM06.01:"Digitalcellulartelecommunicationssystem(Phase2+);Fullratespeech;Processingfunctions".[4]GSM11.10:"Digitalcellulartelecommunicationssystem(Phase2+);MobileStation(MS)conformityspecification".[5]ETS300085:"IntegratedServicesDigitalNetwork(ISDN);3,1kHztelephonyteleservice;Attachmentrequirementsforhandsetterminals(CandidateNET33)".[6]ITU-TRecommendationG.711:"Pulsecodemodulation(PCM)ofvoicefrequencies".[7]ITU-TRecommendationG.712:"Transmissionperformancecharacteristicsofpulsecodemodulation".SIST EN 300 961 V8.0.2:2003
ETSIETSIEN300961V8.0.2(2000-11)9(GSM06.10version8.0.2Release1999)[8]ITU-TRecommendationG.726:"40,32,24,16kbit/sadaptivedifferentialpulsecodemodulation(ADPCM)".[9]ITU-TRecommendationQ.35:"Technicalcharacteristicsoftonesforthetelephoneservice".[10]ITU-TRecommendationV.21:"300bitspersecondduplexmodemstandardizedforuseinthegeneralswitchedtelephonenetwork".[11]ITU-TRecommendationV.23:"600/1200-bandmodemstandardizedforuseinthegeneralswitchedtelephonenetwork".[12]GSM06.32:"Digitalcellulartelecommunicationssystem(Phase2+);VoiceActivityDetector(VAD)".1.1.1AbbreviationsAbbreviationsusedinthepresentdocumentarelistedinGSM01.04.1.2OutlinedescriptionThepresentdocumentisstructuredasfollows:Subclause1.3containsafunctionaldescriptionoftheaudiopartsincludingtheA/DandD/Afunctions.Subclause1.4describestheconversionbetween13bituniformand8bitA-lawsamples.Subclauses1.5and1.6presentasimplifieddescriptionoftheprinciplesoftheRPE-LTPencodinganddecodingprocessrespectively.Insubclause1.7,thesequenceandsubjectiveimportanceofencodedparametersaregiven.Clause2dealswiththetransmissioncharacteristicsoftheaudiopartsthatarerelevantfortheperformanceoftheRPE-LTPcodec.SometransmissioncharacteristicsoftheRPE-LTPcodecarealsospecifiedinclause2.Clause3presentsthefunctionaldescriptionoftheRPE-LTPcodinganddecodingprocedures,whereasclause4describesthecomputationaldetailsofthealgorithm.ProceduresfortheverificationofthecorrectfunctioningoftheRPE-LTParedescribedinclause5.PerformanceandnetworkaspectsoftheRPE-LTPcodecarecontainedinannexA.1.3FunctionaldescriptionofaudiopartsTheanalogue-to-digitalanddigital-to-analogueconversionwillinprinciplecomprisethefollowingelements:1)Analoguetouniformdigital:-microphone;-inputleveladjustmentdevice;-inputanti-aliasingfilter;-sample-holddevicesamplingat8kHz;-analogue-to-uniformdigitalconversionto13bitsrepresentation.Theuniformformatshallberepresentedintwo'scomplement.2)Uniformdigitaltoanalogue:-conversionfrom13bit/8kHzuniformPCMtoanalogue;-aholddevice;-reconstructionfilterincludingx/sinxcorrection;-outputleveladjustmentdevice;SIST EN 300 961 V8.0.2:2003
ETSIETSIEN300961V8.0.2(2000-11)10(GSM06.10version8.0.2Release1999)-earphoneorloudspeaker.Intheterminalequipment,theA/Dfunctionmaybeachievedeither:-bydirectconversionto13bituniformPCMformat;-orbyconversionto8bit/A-orµ-law(PCS1900)compandedformat,basedonastandardA-orµ-law(PCS1900)codec/filteraccordingtoITU-TRecommendationG.711/714,followedbythe8-bitto13-bitconversionaccordingtotheprocedurespecifiedinsubclause1.4.FortheD/Aoperation,theinverseoperationstakeplace.InthelattercaseitshouldbenotedthatthespecificationsinITU-TrecommendationG.714(supersededbyG.712)areconcernedwithPCMequipmentlocatedinthecentralpartsofthenetwork.Whenusedintheterminalequipment,thisspecificationdoesnotonitsownensuresufficientout-of-bandattenuation.Thespecificationofout-of-bandsignalsisdefinedinsection2betweentheacousticsignalandthedigitalinterfacetotakeintoaccountthatthefilteringintheterminalcanbeachievedbothbyelectronicandacousticaldesign.1.4PCMFormatconversionTheconversionbetween8bitA-orµ-law(PCS1900)compandedformatandthe13-bituniformformatshallbeasdefinedinITU-TRecommendationG.721(supersededbyG.726),subclause4.2.1,sub-blockEXPANDandsubclause4.2.7,sub-blockCOMPRESS.TheparameterLAW=1shouldbeusedforA-lawandLAW=0shouldbeusedforµ-law(PCS1900).1.5PrinciplesoftheRPE-LTPencoderAsimplifiedblockdiagramoftheRPE-LTPencoderisshowninfigure1.1.Inthisdiagramthecodingandquantizationfunctionsarenotshownexplicitly.Theinputspeechframe,consistingof160signalsamples(uniform13bitPCMsamples),isfirstpre-processedtoproduceanoffset-freesignal,whichisthensubjectedtoafirstorderpre-emphasisfilter.The160samplesobtainedarethenanalysedtodeterminethecoefficientsfortheshorttermanalysisfilter(LPCanalysis).Theseparametersarethenusedforthefilteringofthesame160samples.Theresultis160samplesoftheshorttermresidualsignal.Thefilterparameters,termedreflectioncoefficients,aretransformedtolog.arearatios,LARs,beforetransmission.Forthefollowingoperations,thespeechframeisdividedinto4sub-frameswith40samplesoftheshorttermresidualsignalineach.Eachsub-frameisprocessedblockwisebythesubsequentfunctionalelements.Beforetheprocessingofeachsub-blockof40shorttermresidualsamples,theparametersofthelongtermanalysisfilter,theLTPlagandtheLTPgain,areestimatedandupdatedintheLTPanalysisblock,onthebasisofthecurrentsub-blockofthepresentandastoredsequenceofthe120previousreconstructedshorttermresidualsamples.Ablockof40longtermresidualsignalsamplesisobtainedbysubtracting40estimatesoftheshorttermresidualsignalfromtheshorttermresidualsignalitself.Theresultingblockof40longtermresidualsamplesisfedtotheRegularPulseExcitationanalysiswhichperformsthebasiccompressionfunctionofthealgorithm.AsaresultoftheRPE-analysis,theblockof40inputlongtermresidualsamplesarerepresentedbyoneof4candidatesub-sequencesof13pulseseach.ThesubsequenceselectedisidentifiedbytheRPEgridposition(M).The13RPEpulsesareencodedusingAdaptivePulseCodeModulation(APCM)withestimationofthesub-blockamplitudewhichistransmittedtothedecoderassideinformation.TheRPEparametersarealsofedtoalocalRPEdecodingandreconstructionmodulewhichproducesablockof40samplesofthequantizedversionofthelongtermresidualsignal.Byaddingthese40quantizedsamplesofthelongtermresidualtothepreviousblockofshorttermresidualsignalestimates,areconstructedversionofthecurrentshorttermresidualsignalisobtained.Theblockofreconstructedshorttermresidualsignalsamplesisthenfedtothelongtermanalysisfilterwhichproducesthenewblockof40shorttermresidualsignalestimatestobeusedforthenextsub-blocktherebycompletingthefeedbackloop.SIST EN 300 961 V8.0.2:2003
ETSIETSIEN300961V8.0.2(2000-11)11(GSM06.10version8.0.2Release1999)1.6PrinciplesoftheRPE-LTPdecoderThesimplifiedblockdiagramoftheRPE-LTPdecoderisshowninfigure1.2.Thedecoderincludesthesamestructureasthefeed-backloopoftheencoder.Inerror-freetransmission,theoutputofthisstagewillbethereconstructedshorttermresidualsamples.Thesesamplesarethenappliedtotheshorttermsynthesisfilterfollowedbythede-emphasisfilterresultinginthereconstructedspeechsignalsamples.1.7SequenceandsubjectiveimportanceofencodedparametersAsindicatedinfigure1.1thethreedifferentgroupsofdataareproducedbytheencoderare:-theshorttermfilterparameters;-theLongTermPrediction(LTP)parameters;-theRPEparameters.Theencoderwillproducethisinformationinauniquesequenceandformat,andthedecodershallreceivethesameinformationinthesameway.Intable1.1,thesequenceofoutputbitsb1tob260andthebitallocationforeachparameterisshown.Thedifferentparametersoftheencodedspeechandtheirindividualbitshaveunequalimportancewithrespecttosubjectivequality.BeforebeingsubmittedtothechannelencodingfunctionthebitshavetoberearrangedinthesequenceofimportanceasgiveninGSM05.03.TherankinghasbeendeterminedbysubjectivetestingandtheprocedureusedisdescribedinannexA,subclauseA.2.Table1.1:Encoderoutputparametersinorderofoccurrenceandbitallocationwithinthespeechframeof260bits/20ms==================================================================ParameterParameterParameterVar.NumberBitno.numbernamenameofbits(LSB-MSB)====================================================================================================================================1LAR16b1-b62LAR26b7-b12FILTER3Log.AreaLAR35b13-b174ratiosLAR45b18-b22PARAMETERS51-8LAR54b23-b266LAR64b27-b307LAR73b31-b338LAR83b34-b36==================================================================Sub-frameno.1==================================================================LTP9LTPlagN17b37-b43PARAMETERS10LTPgainb12b44-b45------------------------------------------------------------------11RPEgridpositionM12b46-b47RPE12BlockamplitudeXmax16b48-b53PARAMETERS13RPE-pulseno.1x1(0)3b54-b5614RPE-pulseno.2x1(1)3b57-b59.25RPE-pulseno.13x1(12)3b90-b92==================================================================SIST EN 300 961 V8.0.2:2003
ETSIETSIEN300961V8.0.2(2000-11)12(GSM06.10version8.0.2Release1999)Sub-frameno.2==================================================================LTP26LTPlagN27b93-b99PARAMETERS27LTPgainb22b100-b101------------------------------------------------------------------28RPEgridpositionM22b102-b103RPE29BlockamplitudeXmax26b104-b109PARAMETERS30RPE-pulseno.1x2(0)3b110-b11231RPE-pulseno.2x2(1)3b113-b115.42RPE-pulseno.13x2(12)3b146-b148==================================================================Sub-frameno.3==================================================================LTP43LTPlagN37b149-b155PARAMETERS44LTPgainb32b156-b157------------------------------------------------------------------45RPEgridpositionM32b158-b159RPE46BlockamplitudeXmax36b160-b165PARAMETERS47RPE-pulseno.1x3(0)3b166-b16848RPE-pulseno.2x3(1)3b169-b171.59RPE-pulseno.13x3(12)3b202-b204==================================================================Sub-frameno.4==================================================================LTP60LTPlagN47b205-b211PARAMETERS61LTPgainb42b212-b213------------------------------------------------------------------62RPEgridpositionM42b214-b215RPE63BlockamplitudeXmax46b216-b221PARAMETERS64RPE-pulseno.1x4(0)3b222-b22465RPE-pulseno.2x4(1)3b225-b227.76RPE-pulseno.13x4(12)3b258-b260==================================================================SIST EN 300 961 V8.0.2:2003
ETSIETSIEN300961V8.0.2(2000-11)13(GSM06.10version8.0.2Release1999)InputPre-processingsignalShorttermanalysisfilterShorttermLPCanalysis+RPEgridselectionandcoding(1)(2)LTPanalysisLongtermanalysisfilter+RPEgriddecodingandpositioning(4)(5)(3)-LTPparameters(9bits/5ms)ReflectioncoefficientscodedasLog.-AreaRatios(36bits/20ms)RPEparameters(47bits/5ms)Toradiosubsystem(1)Shorttermresidual(2)Longtermresidual(40samples)(3)Shorttermresidualestimate(40samples)(4)Reconstructedshorttermresidual(40samples)(5)Quantizedlongtermresidual(40samples)Figure1.1:SimplifiedblockdiagramoftheRPE-LTPencoderRPEgriddecodingandpositioningReflectioncoefficientscodedasLog.-AreaRatios(36bits/20ms)RPEparameters(47bits/5ms)FromradiosubsystemLTPparameters(9bits/5ms)+ShorttermsynthesisfilterLongtermsynthesisfilterPost-processingOutputsignalFigure1.2:SimplifiedblockdiagramoftheRPE-LTPdecoderSIST EN 300 961 V8.0.2:2003
ETSIETSIEN300961V8.0.2(2000-11)14(GSM06.10version8.0.2Release1999)2TransmissioncharacteristicsThisclausespecifiesthenecessaryperformancecharacteristicsoftheaudiopartsforproperfunctioningofthespeechtranscoder.SometransmissionperformancecharacteristicsoftheRPE-LTPtranscoderarealsogiventoassistthedesignerofthespeechtranscoderfunction.TheinformationgivenhereisredundantandthedetailedspecificationsarecontainedinrecommendationGSM11.10.Theperformancecharacteristicsarereferredtothe13bituniformPCMinterface.NOTE:Tosimplifytheverificationofthespecifications,theperformancelimitsmaybereferredtoanA-orµ-law(PCS1900)measurementinterfaceaccordingtoITU-TRecommendationG.711.Inthisway,standardmeasuringequipmentsforPCMsystemscanbeutilizedformeasurements.Therelationshipbetweenthe13bitformatandtheA-orµ-law(PCS1900)compandedshallfollowtheproceduresdefinedinsubclause1.4.2.1Performancecharacteristicsoftheanalogue/digitalinterfacesConcerning1)discriminationagainstout-of-bandsignals(sending)and2)spuriousout-of-bandsignals(receiving),thesamerequirementsasdefinedinETSIstandardTE04-15(digitaltelephone,candidateNET33)apply.2.2TranscoderdelayConsiderabacktobackconfigurationwheretheparametersgeneratedbytheencoderaredeliveredtothespeechdecoderassoonastheyareavailable.Thetranscoderdelayisdefinedasthetimeintervalbetweentheinstantaspeechframeof160sampleshasbeenreceivedattheencoderinputandtheinstantthecorresponding160reconstructedspeechsampleshavebeenout-putbythespeechdecoderatan8kHzsamplerate.Thetheoreticalminimumdelaywhichcanbeachievedis20ms.Therequirementisthatthetranscoderdelayshouldbelessthan30ms.3FunctionaldescriptionoftheRPE-LTPcodecTheblockdiagramoftheRPE-LTP-coderisshowninfigure3.1.Theindividualblocksaredescribedinthefollowingsubclauses.3.1FunctionaldescriptionoftheRPE-LTPencoderThePre-processingsectionoftheRPE-LTPencodercomprisesthefollowingtwosub-blocks:-offsetcompensation(3.1.1);-pre-emphasis(3.1.2).TheLPCanalysissectionoftheRPE-LTPencodercomprisesthefollowingfivesub-blocks:-segmentation(3.1.3);-auto-Correlation(3.1.4);-schurRecursion(3.1.5);-transformationofreflectioncoefficientstoLog.-AreaRatios(3.1.6);-quantizationandcodingofLog.-AreaRatios(3.1.7).SIST EN 300 961 V8.0.2:2003
ETSIETSIEN300961V8.0.2(2000-11)15(GSM06.10version8.0.2Release1999)TheShorttermanalysisfilteringsectionoftheRPE-LTPcomprisesthefollowingfoursub-blocks:-decodingofthequantizedLog.-AreaRatios(LARs)(3.1.8);-interpolationofLog.-AreaRatios(3.1.9);-transformationofLog.-AreaRatiosintoreflectioncoefficients(3.1.10);-shorttermanalysisfiltering(3.1.11).TheLongTermPredictor(LTP)sectioncomprises4sub-blocksworkingonsubsegments(3.1.12)oftheshorttermresidualsamples:-calculationofLTPparameters(3.1.13);-codingoftheLTPlags(3.1.14)andtheLTPgains(3.1.15);-decodingoftheLTPlags(3.1.14)andtheLTPgains(3.1.15);-longtermanalysisfiltering(3.1.16),andLongtermsynthesisfiltering(3.1.17).TheRPEencodingsectioncomprisesfivedifferentsub-blocks:-weightingfilter(3.1.18);-adaptivesampleratedecimationbyRPEgridselection(3.1.19);-APCMquantizationoftheselectedRPEsequence(3.1.20);-APCMinversequantization(3.1.21);-RPEgridpositioning(3.1.22).Pre-processingsection3.1.1OffsetcompensationPriortothespeechencoderanoffsetcompensation,byanotchfilterisappliedinordertoremovetheoffsetoftheinputsignalsotoproducetheoffset-freesignalsof.sof(k)=so(k)-so(k-1)+alpha*sof(k-1)(3.1.1)alpha=32735*2-153.1.2Pre-emphasisThesignalsofisappliedtoafirstorderFIRpre-emphasisfilterleadingtotheinputsignalsoftheanalysissection.s(k)=sof(k)-beta*sof(k-1)(3.1.2)beta=28180*2-15LPCanalysissection3.1.3SegmentationThespeechsignals(k)isdividedintonon-overlappingframeshavingalengthofT0=20ms(160samples).AnewLPC-analysisoforderp=8isperformedforeachframe.SIST EN 300 961 V8.0.2:2003
ETSIETSIEN300961V8.0.2(2000-11)16(GSM06.10version8.0.2Release1999)3.1.4AutocorrelationThefirstp+1=9valuesoftheAuto-Correlationfunctionarecalculatedby:159ACF(k)=s(i)s(i-k),k=0,1.,8(3.2)i=k3.1.5SchurRecursionThereflectioncoefficientsarecalculatedasshowninfigure3.2usingtheSchurRecursionalgorithm.Theterm"reflectioncoefficient"comesfromthetheoryoflinearpredictionofspeech(LPC),whereavocaltractrepresentationconsistingofseriesofuniformcylindricalsectionsisassumed.Sucharepresentationcanbedescribedbythereflectioncoefficientsorthearearatiosofconnectedsections.3.1.6TransformationofreflectioncoefficientstoLog.-AreaRatiosThereflectioncoefficientsr(i),(i=1.8),calculatedbytheSchuralgorithm,areintherange:-1<=r(i)<=+1Duetothefavourablequantizationcharacteristics,thereflectioncoefficientsareconvertedintoLog.-AreaRatioswhicharestrictlydefinedasfollows:1+r(i)Logarea(i)=log10(----------)(3.3)1-r(i)Sinceitisthecompandingcharacteristicofthistransformationthatisofimportance,thefollowingsegmentedapproximationisused.r(i);|r(i)|<0.675LAR(i)=sign[r(i)]*[2|r(i)|-0.675];0.675<=|r(i)|<0.950sign[r(i)]*[8|r(i)|-6.375];0.950<=|r(i)|<=1.000(3.4)withtheresultthatinsteadofhavingtodivideandobtainthelogarithmofparticularvalues,itismerelynecessarytomultiply,addandcomparethesevalues.Thefollowingequation(3.5)givestheinversetransformation.LAR'(i);|LAR'(i)|<0.675r'(i)=sign[LAR'(i)]*[0.500*|LAR'(i)|+0.337500];0.675<=|LAR'(i)|<1.225sign[LAR'(i)]*[0.125*|LAR'(i)|+0.796875];1.225<=|LAR'(i)|<=1.625(3.5)3.1.7QuantizationandcodingofLog.-AreaRatiosTheLog.-AreaRatiosLAR(i)havedifferentdynamicrangesanddifferentasymmetricdistributiondensities.Forthisreason,thetransformedcoefficientsLAR(i)arelimitedandquantizeddifferentlyaccordingtothefollowingequation(3.6),withLARc(i)denotingthequantizedandintegercodedversionofLAR(i).LARc(i)=Nint{A(i)*LAR(i)+B(i)}(3.6)withNint{z}=int{z+sign{z}*0.5}(3.6a)FunctionNintdefinestheroundingtothenearestintegervalue,withthecoefficientsA(i),B(i),anddifferentextremevaluesofLARc(i)foreachcoefficientLAR(i)givenintable3.1.SIST EN 300 961 V8.0.2:2003
ETSIETSIEN300961V8.0.2(2000-11)17(GSM06.10version8.0.2Release1999)Table3.1:QuantizationoftheLog.-AreaRatiosLAR(i)LARNoiA(i)B(i)MinimumLARc(i)MaximumLARc(i)120.0000.000-32+31220.0000.000-32+31320.0004.000-16+15420.000-5.000-16+15513.6370.184-8+7615.000-3.500-8+778.334-0.666-4+388.824-2.235-4+3Short-termanalysisfilteringsectionThecurrentframeofthespeechsignalsisretainedinmemoryuntilcalculationoftheLPCparametersLAR(i)iscompleted.Theframeisthenreadoutandfedtotheshorttermanalysisfilteroforderp=8.However,priortotheanalysisfilteringoperation,thefiltercoefficientsaredecodedandpre-processedbyinterpolation.3.1.8DecodingofthequantizedLog.-AreaRatiosInthisblockthequantizedandcodedLog.-AreaRatios(LARc(i))aredecodedaccordingtoequation(3.7).LAR''(i)=(LARc(i)-B(i))/A(i)(3.7)3.1.9InterpolationofLog.-AreaRatiosToavoidspurioustransientswhichmayoccurifthefiltercoefficientsarechangedabruptly,twosubsequentsetsofLog.-AreaRatiosareinterpolatedlinearly.Withineachframeof160analysedspeechsamplestheshorttermanalysisfilterandtheshorttermsynthesisfilteroperatewithfourdifferentsetsofcoefficientsderivedaccordingtotable3.2.Table3.2:InterpolationofLARparameters(J=actualsegment)kLAR'J(i)=0.120.75*LAR''J-1(i)+0.25*LAR''J(i)13.260.50*LAR''J-1(i)+0.50*LAR''J(i)27.390.25*LAR''J-1(i)+0.75*LAR''J(i)40.159LAR''J(i)3.1.10TransformationofLog.-AreaRatiosintoreflectioncoefficientsThereflectioncoefficientsarefinallydeterminedusingtheinversetransformationaccordingtoequation(3.5).3.1.11ShorttermanalysisfilteringTheShorttermanalysisfilterisimplementedaccordingtothelatticestructuredepictedinfigure3.3.d0(k)=s(k)(3.8a)u0(k)=s(k)(3.8b)di(k)=di-1(k)+r'i*ui-1(k-1)withi=1,.8(3.8c)ui(k)=ui-1(k-1)+r'i*di-1(k)withi=1,.8(3.8d)d(k)=d8(k)(3.8e)Long-TermPredictor(LTP)sectionSIST EN 300 961 V8.0.2:2003
ETSIETSIEN300961V8.0.2(2000-11)18(GSM06.10version8.0.2Release1999)3.1.12Sub-segmentationEachinputframeoftheshorttermresidualsignalcontains160samples,correspondingto20ms.Thelongtermcorrelationisevaluatedfourtimesperframe,foreach5mssubsegment.Forconvenienceinthefollowing,wenotej=0,.,3thesub-segmentnumber,sothatthesamplespertainingtothej-thsub-segmentoftheresidualsignalarenowdenotedbyd(kj+k)withj=0,.,3;kj=k0+j*40andk=0,.,39wherek0correspondstothefirstvalueofthecurrentframe.3.1.13CalculationoftheLTPparametersForeachofthefoursub-segmentsalongtermcorrelationlagNj,(j=0,.,3),andanassociatedgainfactorbj,(j=0,.,3)aredetermined.Foreachsub-segment,thedeterminationoftheseparametersisimplementedinthreesteps.1)Thefirststepistheevaluationofthecross-correlationRj(lambda)ofthecurrentsub-segmentofshorttermresidualsignald(kj+i),(i=0,.,39)andtheprevioussamplesofthereconstructedshorttermresidualsignald'(kj+i),(i=-120,.,-1):39j=0,.3Rj(lambda)=d(kj+i)*d'(kj+i-lambda);kj=k0+j*40i=0lambda=40,.,120(3.9)Thecross-correlationisevaluatedforlagslambdagreaterthanorequalto40andlessthanorequalto120,i.e.correspondingtosamplesoutsidethecurrentsub-segmentandnotdelayedbymorethantwosub-segments.2)ThesecondstepistofindthepositionNjofthepeakofthecross-correlationfunctionwithinthisinterval:Rj(Nj)=max{Rj(lambda);lambda=40.120};j=0,.,3(3.10)3)Thethirdstepistheevaluationofthegainfactorbjaccordingto:bj=Rj(Nj)/Sj(Nj);j=0,.,3(3.11)with39Sj(Nj)=d'2(kj+i-Nj);j=0,.,3(3.12)i=0Itisclearthatthelast120samplesofthereconstructedshorttermresidualsignald'(kj+i),(i=-120,.,-1)shallberetaineduntilthenextsub-segmentsoastoallowtheevaluationoftherelations(3.9),.,(3.12).3.1.14Coding/DecodingoftheLTPlagsThelongtermcorrelationlagsNj,(j=0,.,3)canhavevaluesintherange(40,.,120),andsoshallbecodedusing7bitswith:Ncj=Nj;j=0,.,3(3.13)Atthereceivingend,assuminganerrorfreetransmission,thedecodingofthesevalueswillrestoretheactuallags:Nj'=Ncj;j=0,.,3(3.14)SIST EN 300 961 V8.0.2:2003
ETSIETSIEN300961V8.0.2(2000-11)19(GSM06.10version8.0.2Release1999)3.1.15Coding/DecodingoftheLTPgainsThelongtermpredictiongainsbj,(j=0,.,3)areencodedwith2bitseach,accordingtothefollowingalgorithm:ifbj<=DLB(i)thenbcj=0;i=0ifDLB(i-1)
...