基于小波變換的語音信號(hào)的噪聲處理研究.pdf

1、l(】80j04廿璺0鞠嘶蟲害研究生學(xué)位論文嶄蛐論史地I『I“1t文)立IU：：n超生拯也匿童盟上姐蠑已『地業(yè)存嘭‘一一噸芷地(外文)——一一墨l砣!!!!墨ig旦!l旦墊j!it一_In一＼Viwelet盟￡!h堅(jiān)I!州究生姓名一一堂弛瞧一與!：科、專業(yè)——一道信與信丘j!亟塹研究方向——道信垂贊熬娥I，i矬埋帥姓私、職稱——一一～一崮亞韭刨熬絲一鹼艾『件。培El：年月一一一3壁堂生璺』]至：業(yè)2鹽5』J泛殳提受[I期一!!盟王E業(yè)詭

2、殳答辯口蚓————～土盟一蔓；!j一——蘭州大學(xué)研究生學(xué)位論文AbstractSpeechsignalde—noisingisanimportantcomponentinthesignalprocessingWaveletcananalyzesignalattimedomainandfrequencysimultaneouslySOitcandenoiseeffectivelyBecausethemotherwaveletsiSagat

3、herwhichmotherwaveletCanreachbestdenoisingeffectatpracticalapplicationwilltoberesolveThentheelectionofthresholdshasimmediaterelationtotheresultofde—noisingSomewaveletcoefficientscantbesetzerowhenthethresholdisundersizean

4、dpartsofnoisesareretainedSomeusefulsignalswillbereducedifthethresholdisupsizeThesecasesmaydegradethedenoisingeffectTwoquestionsaremainlystudiedinthisthesisInthisarticlefivelayersofdb3waveletisselectedasmotherwavelet，four

5、thresholdselectionrulesanddifferentrescalingmethodsareusedtoanalysisthenoisyspeechsignalAnovelthresholddenoisingmethodnamedadaptivewaveletcoefficientssubtractionisproposedbasedonshorttimeFouriertransformation111eproposed

6、methodisbasedonSUREunbiasestimationandthemostoptimalthreshholdparameterscanberesearchedadaptivelysatisfytherequirementoftimevariablesignalinreal—timeprocessingAtlast，theproposedmethodissimulatedusingMatlab70software，andt

7、heresultsarecomparedwiththatofspectrumsubtractivemethodTheresultsofsimulationprovethatadaptivewaveletcoefficientssubtractionmethodCalleffectivelyrestrainnoisesandimproveSNRdramaticlyKeywords：WaveletAnalysis；De—noising；Mo