基于ICA的盲信源分離.pdf

1、西安電子科技大學(xué)碩士學(xué)位論文基于ICA的盲信源分離姓名：劉小東申請學(xué)位級別：碩士專業(yè)：模式識別與智能系統(tǒng)指導(dǎo)教師：樓順天20030101AbstractTheblindsourcesepararon(Bss)whentheSOUrCenumberisunknownandchangesdynamicallyisallimportantandopenproblem，Inrecentyears，manyseparatingalgorithms

2、havebeendeveloped，inwhichsomealgorithmssuchasthenaturalgradientalgorithmandrelativegradientalgorithmworkeffectiveHowevermestofalgorithmsrequirethatthenumberofsourcesisassumedtobeknownaprioriandunchangedInotherword，thosea

3、lgorithmscanriotalwaysobtaingoodseparationresults，whenthesourcenumberisunknownandchangesdynamicallyAnewalgorithmisintroducedinthispaperwhichisapplied船thedeterminationofunknownnumberofsourcesandthedetectionofthesourcenumb

4、erdynamicallyaswd!astheontineblindseparationofoTarce$Thekeypointsofthepaperaleasfollows：BasedonthecorrelationcoefficiemamongtheelementsOfoutptRvectorofneuralnetwork，theredundantsignaiinoutputcomponentscanbedetected，andde

5、letedbychangingthestructureofdynamicneuralnetwork(DNN)，TheDNNusingaatu罐gradientalgorithmcanbechange4itsstructureswhenthesourconumbercartischangedwithDNN，theunknownsourcenumbercanbedetermined，thechangingsourcenumbercallbe

6、detectedandthemixedsourceCallbeseparatedonlineBased0ntheblindsourceseparationwithindependentcomponentanalysis(ICA)andwavdetpacketdecomposition鑫newalgorithmforthedeterminationoftheSOUSenumberispresentedusingtheenergychara

7、cteristicvectorInthecaseofunknownordynamicallychangingnumb_erofsources，themixedsouf輯isseparatedonlineusing[CAwiththenaturalgradientalgorithm，Thenbasedonthewaveletpacketdecompositionofoutputcomponent，theenergycharacterist

8、icvec治rofeachoutputcomponentisobtainedAfter氌揀thedistanceamongtheelementsofenergycharacteristicvectorarecalculatedonlineThecorrelationoutputsignalcomponentscallbedetectedb豁酣onthesedistancesSuchthemqknownnumberofsourcecanb