低碼率特征波形內(nèi)插語音編碼算法的關(guān)鍵技術(shù)研究.pdf

1、東南大學(xué)碩士學(xué)位論文低碼率特征波形內(nèi)插語音編碼算法的關(guān)鍵技術(shù)研究姓名：黃蓉申請(qǐng)學(xué)位級(jí)別：碩士專業(yè)：信號(hào)與信息處理指導(dǎo)教師：吳鎮(zhèn)揚(yáng)20090101AbstractAbstract11埔lowratespeechcodingisthecrucialtechnologyofpresentcommunicationfieldOnthefoundationoftheCharacteristicWaveformInterpolation(CⅥ)sp

2、eechcodingalgorithm，thispap明“presentsakindofcodingalgorithmof18kbit／sCWIbased01“Iphonemesegmentationandthecodingalgorithmof24kbit／sCWIbasedonmulfibandThemajorresearchaccomplishmentsofthispaperinclude：1Thespeechsegmentati

3、onalgorithminthelowbitratespeechcodingisstudiedAkindofspeechsegmentationalgorithmbasedOnthecumulateenergyofwaveletuansformationisproposedanditpromotestheprecisionofspeechsegmentationTheeffectivedivisionofvoicedsegmentand

4、unvoicedsegmentisdonethroughspeechsegmentation，andonlythevoicedsegmentiscodedThecorrelationoftl鴝speechsignalofeachspeechsegmentationisusedtocompressandcodetheparameter2ThequantizationmethodofLSFiSstudiedConsideringthefTa

5、l_11ccorrelationofLSFparameters。aquantizationmethodofLSFbasedOntemporaldecompositionisstudiedThecodingspeedofLSFisreducedto695bit／s，andthecodingspeedisgreatlyreducedandarelativelylowerspectraldistortioninaverageisgotten3

6、田地quantizationmethodofcharacteristicwaveformisstudiedToraisecodingefficiencycharacteristicwaveformisdecomposedintoSEWandREW，whicha∞quantifiedrespectivelyaccordingtotheirdifferentperceptualcharacteristicForSEW，quantizatio

7、nisdonebydividingitsamplitudespectrumintothreesubbandForPEW，orthogonalpolynomialfittingisused4A1，8kbit／sCWIspeechcodingalgorithmbasedonphonemesegmentationisproposedThisalgorithmUSeSphonemesegmentationasfrontendprocessing

8、，andquantifiesLSFparametersbytemporaldeI舳mpositionalgorithmItalsoconsiderstheintervalofeventlocationtObethefinalspeechsegmentationpointandquantifiesparametersonthefoundationofsegmentationSubjectivequalitytestsshowthatsyn

9、theticspeechmaintainshigherintelligibilityandcertainllaturaldegree，butitsqualityisslightlylowerthanthealgorithmof24kbit／sMELP5A24kbit／sCWIspeechcodingalgorithmbasedonmultibandisproposedThisalgorithmUSeStheideaofmultiband

10、intheMELPcodingalgorithmanddividesthespeechspectrumintoseveralfrequencybands，inwhichtoanalyzethevoicestrengthAccordingtOtheanalyzingresult，thephasesofSEWandPEWintherespectivefxequencybandaredeterminedandthereforetheMghha

11、rmonicsofthespeechCanberecoveredwellMeanwhile，thecharacteristicwaveformsarepartlyalignedwheninterpolatingtheinstantaneouswaveformssothatthecomputingcomplexityisgreatlyreducedSubjectivequalitytestsRA／Bshowsthatthequalityo

12、fsynthesizespeechisgreatbetterthanthetypical24kbitYsCWIcodingalgorithmandslightlybetterthanthe24kbit／sMELPcodingalgorithmKeyWords：lowratespeechcoding；speechsegmentation；temporaldecomposition；characteristicwaveform；multi—