资源描述：
《压缩感知测量优化分析》由会员上传分享，免费在线阅读，更多相关内容在工程资料-天天文库。
1、AbstractCompressedSensingtheoryisanewcodecmethodwhichwasproposedinrecentyears.ItbreakstheNyquistsamplingtheoremconstraints.Comparedtothetraditionalcodectheory,thegreatestadvantageofcompressedsensingcodectheoryisitsencodingprocessisextremelysimple,thecomplexityistransferredtothedecodingsi
2、de,whichhasitsuniqueadvantagesinstrongmobility,limitedcomputingpowerandlowerstoragecapacityoccasion.Butcomparedtothetraditionalcodectheory,signalreconstructionqualityofcompressedsensingcodecmethodremainstobeimproved.Themainpurposeofthispaperistooptimizethecodingmethodofcompressedsensing,
3、includingoptimizationofthemeasurementmatrix,adaptiveselectionofmeasurements,measurementsquantizationmethod.Asthecodingmethodofcompressedsensingisaprojectionprocessofthesignalontoameasurementmatrix,thequalityofmeasurementmatrixdirectlyaffectsthecodecquality,sothepaperstartsfromoptimizingo
4、fthemeasurementmatrix,Twomeasurementmatrixoptimizationalgorithmsisproposed:oneistheuppertriangularweightedmeasurementmatrixoptimizationalgorithm,whichcanenhancethesamplingoflowfrequencycoefficients;theotheristhegradient-basedGrammatrixiterativeoptimization,whichcanreducetherelevanceofmea
5、surementmatrixandsparsematrix.Experimentsshowthatthesetwooptimizationscanimprovetheperformanceofexistingmeasurementmatrix.Secondly,amaximumposteriorivariancebasedadaptiveselectionofmeasurementsalgorithmisproposed,whichismainlyaboutMForaknownmeasurementmatrix,selectingwhichlinescangettheb
6、estmeasurements11and"Selectinghowmanylinesareadequateforthecurrentsignal11.Accordingtothisalgorithm,ameasurementmatrixwhichhaslargenumberrowsisselectedfirst,thenselectingtherightrowsaccordingtothisalgorithm.Experimentalresultsshowthattheproposedadaptivealgorithmcangetbetterreconstruction
7、qualitycomparedtotheconventionalmeasurementmethodatthesameconditions.Finally,researchaboutthequantizationofcompressedsensingmeasurementsisdoneinthispape匚Worksaboutquantizationofmeasurementsisnottoomuch,butinpracticalapplications,quantizationmustbeconsidered.Asthetheoryofc
