基于热释电红外信息的人体身份识别的研究

《基于热释电红外信息的人体身份识别的研究》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。

1、ABSTRACTGaitrecognitionisanewfieldfor．thebiometricrecognitiontechnology．Itsaimistorecognizepeopleordetectphysiological，pathologicalandmentalcharactersbytheirwalkingposture．Therefore，thistechniquehasabrightprospectandattractsmoreandmoreattentionfromthefieldofbiomedi

2、calinformationdetection．Pyroelectricinfrared(PIR)sensorscandetecthumanbodymotionwithinafairlyreasonabledistancebytheirheatemissions．PIRsensorshavebeenusedforawiderangeofapplications，suchasintruderdetectionandlightactuatorsduetoitslowcostandlowpowerconsumption．Thein

3、fraredradiationofawalkingpersonalsocontainsphysicalcharacteristicsoflimbswingandwalkingpostureinformation，whichprovidesanewsourceforgaitfeatureextractionandidentification．HumanbodymovementcharacteristicextractionforawalkingpersonisinvestigatedinthisthesisusingaPIRs

4、ensorwithFresneilensarrays．TemporalvoltagesignalscollectedfromaPIRsensorwereanalyzedtoextractthemotionfeaturesofindividualsforhumanidentification．Inourstudy,wesetupaninfrareddataacquisitionsystem，conducteddataacquisitionexperimentandestablishedasmall·scaleinfraredd

5、atabase．Thentwomethodswereusedforextractingthefeaturesforidentification．Ontheonehand，weestablishedauto-regressive(AR)modelsfortime-domainsignals，extractedARcoefficientsasthecharacteristicsandthenusedsupportvectormachine(SVM)fortheclassificationprocess．Thehighestpro

6、babilityofcorrectrecognitionwas66．48％：Ontheotherhand．weextractedspectruminformationofthetime—domainsignalbyusingFourierTransform，andemployedprincipalcomponentanalysis(PCA)methodfordimensionalityreductionofspectrummatrix．Thehighestprobabilityofcorrectrecognitionwas8

7、6．53％inSVMclassifier．Theexperimentalresultsshowedthatthespectruminformationofpyroelectricinfraredsignalbytime—frequencytransformingfollowedbyprocessingofPCA，wasbetterthanthefirstmethodoffeatureextraction．Theresearchdisplayedthatidentificationofhumanbodybyprocessing

8、oftheoutputsignalofaPIRsensorcanbeachievedtosomeextent，andthisWasalsolow-costhumanidentificationsystemprogramsinsomelow-securityapplications．Atpr