電子鼻識(shí)別中的預(yù)處理與模糊神經(jīng)網(wǎng)絡(luò)算法.pdf

1、大連理工大學(xué)碩士學(xué)位論文電子鼻識(shí)別中的預(yù)處理與模糊神經(jīng)網(wǎng)絡(luò)算法姓名：楊俊申請(qǐng)學(xué)位級(jí)別：碩士專業(yè)：物理電子學(xué)指導(dǎo)教師：閆衛(wèi)平20071220大連理1=大學(xué)碩十學(xué)位論文ArithmeticofSignalPreprocessingandFuzzyNeuralNetworkinRecognitionofElectronicNoseAbstractCausedbythephysicalshortcomingsofgassL=Tlsors，iti

2、simpossibleforasinglegassL_flSOrtoidentifymultiplegasesSotheelectronicnosetechniquesbasedongassensorarrayandpatternrecognitionisbecominganimportantwayindealingwithcross—sensitivityingasanalysisThepatternrecognitiontechno

3、logyplaysthecrucialroletothebehaviorofelectronicnoseAimedtothisrequirement，theresearchstudiesseveralrelatedpatternrecognitiontechniquesinordertealizethequantitativelyrecognitionofthemixedgasesTheresearchstudiesthePrincip

4、alComponentsAnalysisandIndependentComponentAnalysisarithmeticTheprincipalcomponentanalysiscouldsubstituteasmallnumbercomprehensivevariablefororiginalmultidimensionalvariableandobviouslysimplifydatastructureinthepre—condi

5、tionofminimizinglossoforiginaldatainformationTheresearchcombinesthePrincipalComponentsAnalysisandIndependentComponentAnalysisinordertealizeaneffectivedatapre—processingintheelectronicnosesystemSomeexperimentsareperformed

6、toverifytheoptimizingeffectofPCAandICAAimedtotheneuralnetworkandthefuzzy109icofthepatternrecognitiontechnique，combinestheIogicalexpressionabilityofthefuzzysystemandtheselfstudingabilityoftheneuralnetwork，bringsforwardafu

7、zzyneuralnetworkarithmetic：Takagi—SugenofuzzyneuralnetworktorealizethequantitativelyrecognitionofthealcoholthicknessBasedonthismethod，bringsforwardanadvancedTakagi—SugenoFuzzyLogicSystembasedonNeuralNetworksitusestheneur

8、alnetworktosimulatethesubjectiveselectionThismethodCanavoidthelowefficiencyandimpreciseofthesubjectiveselectionandobtaintheassociatedmembershipfunctionandreasoningregulationaccuratelyTheresearchusesthealcoholelectronicno

9、setestingdateastheswatchusingthearithmeticmentionedabovetosimulateandtest54kindsofmixedgasestheaccuraterateofthequantitativelyrecognitionCanachievetoabove99％111eresultprovesthatthesignalpreprocessingtechniquesandthefuzzy