基于聲學(xué)與小波熵及自回歸模型的病態(tài)嗓音診斷新方法研究.pdf

1、IIlllllllllll_|ll!ll!lll!11111111111llllllllllllllIY1531769天津大學(xué)申請(qǐng)博士學(xué)位論文基于聲學(xué)與小波熵及自回歸模型的病態(tài)嗓音診斷新方法研究StudyontheNovelMethodofPathologicalVoiceDiagnosisBasedonAcoustics，WaveletEntropyandAuto—Regressivemodel一級(jí)學(xué)科：學(xué)科專(zhuān)業(yè)：作者姓名：指導(dǎo)教師：

2、生塑醫(yī)堂工程生翅匡堂王捏蒸筮互趨壟夔握天津大學(xué)精密儀器與光電子工程學(xué)院2008年1月9曰ABSTRACTThecomputerizedpathologicalvoiceevaluationanddiagnosisisanewfieldfortheearlytherapy，illnessdetectionandmedicaltreatmentinformatizationtechnologySinceitsobjective，rapida

3、ndgeneralusedcharacteritisbecomingtheimportantandassistantmeasurementtOpreventtheealydiagnosesomefataldiseasesThenovelmethodofpathologicalvoiceevaluationanddiagnosismentionedinthispaperisatoolofearlydetectionanddiagnosis

4、ofillnesswichistobeconcemedThisareafallsbackonthebasesofacousticsandanatomiatheorybymeansofcomputerandmordendigitalsignalprocessingtechniquetOanalyzethesoundsampledataintimeorspectrumdomainandclassifytheminmultidimension

5、alfeaturesspaceIntheresearchthedatabaseissuppliedbyMasachusettsEyeandEarInfumaryThemajorresearchisincludingtheextractionsof19aouctics，5waveletentropyand13AutoRegressivemodelfeaturesstraightlyfromthevoicesamplesThesupport

6、vectormachineisemployedastheclassifierinbi—classificationofhealthyandpathologyvoicewiththedifferentfeaturegroupofabovethethreeAndthenthefurthertaskishyperfunctionandfolddamageillnessdiscriminationAtlastdependinonthefinal

7、expermentsresult，thehyperfunctionandfolddamagedillnessspecializedfeaturesarefoundedamongthesegroups’featuresbyusingtheorthogonalexperimentdesignTheexpermentseffectsshowthatacousticsandAuto—Regressivemodelgoroup’Sfeatures

8、areappearancesobviousinhealthyandpathologydivisionTheaccuraciesare897％and878％respectivelyThoughttheAuto—Regessivemodelfeaturesarefirstused，itnotonlywelldoneinpathologicalvoicedetectionbutalsohasabestobviuosperformenancea

9、ndmuchbetterthantheothertwogroupsfeaturesinpathologicalvoiceidentificationwithaccuracyof873％The1髓to5m8血to9mand11血ARmodelcoefficientsarechoosenanddeterminedtOhavespecializedcharacteristiesTheinnovationsofthisthesisarethef

10、ollowings：iThefnstcombinationusingofacoustics，waveletentropyandAuto—RegressivemodelfeaturesinresearchofpathologicalvoicediagnosisItisthegreatexpandationworktotheusingfeaturessetandbenificaltoillnessrealystagediagnosisiiF