简介:艾丽斯识别享受普遍性,唯一的高度和中等用户合作。这在新兴的安全与认证机制使虹识别系统不可避免。一个虹识别系统基于向量量子化(VQ),技术被建议,它的表演与分离余弦变换(DCT)相比。建议系统不需要虹的任何预处理和分割。我们测试了林德布佐·格雷(LBG),Kekre的相称的错误(KPE)算法和Kekre是为聚类的目的快电报密码本产生(KFCG)算法。基于的方法使用KFCG要求的建议向量量子化99.99%更少的计算作为完整的2-dimensionalDCT的。进一步,KFCG方法与89.10%的精确性给更好的性能超过在66.10%附近给精确性的DCT。
简介:IntheH.263videocodecrelatedsystems,motionestimationandDiscreteCosineTransform(DCT)havethemostcomputationalrequirements.Inordertoreducecomplexityoftheencodertodedicatemoreresourcestootherfunctions,accordingtothestudyofexistingmethods,anImprovedAllZeroBlockFinding(IAZBF)methodbasedonthestatisticcharacteristicsofDCTcoefficientsisproposed.Comparedwithexistingmethods,IAZBFimprovesthedetectingefficiencybyabout50%withoutimportingtoomuchextracomputationrequirement.Beingcomputedwithadditionsandshiftsinsteadofcomplicatedmultiplications,IAZBFisoflowcomputationcomplexity,especiallyforlow-endprocessors.Inaddition,IAZBFupholdspicturefidelityandremainscompatiblewiththeH.263bitstreamstandard.
简介:Tocompresshyperspectralimages,alowcomplexitydiscretecosinetransform(DCT)-baseddistributedsourcecoding(DSC)schemewithGraycodeisproposed.UnlikemostoftheexistingDSCschemes,whichutilizetransforminspatialdomain,theproposedalgorithmappliestransforminspectraldomain.Set-partitioning-basedapproachisappliedtoreorganizeDCTcoefficientsintowaveletliketreestructureandextractthesign,refinement,andsignificancebitplanes.TheextractedrefinementbitsareGrayencoded.Becauseofthedependencyalongthelinedimensionofhyperspectralimages,lowdensityparitycheck-(LDPC)-basedSlepian-WolfcoderisadoptedtoimplementtheDSCstrategy.Experimentalresultsonairbornevisible/infraredimagingspectrometer(AVIRIS)datasetshowthattheproposedparadigmachievesupto6dBimprovementoverDSC-basedcoderswhichapplytransforminspatialdomain,withsignificantlyreducedcomputationalcomplexityandmemorystorage.
简介:Inthispaper,twospeechenhancementsystemswithsupergaussianspeechmodelingarepresented.ThecleanspeechcomponentsareestimatedbyMinimum-Mean-Square-Error(MMSE)es-timatorundertheassumptionthattheDCTcoefficientsofcleanspeecharemodeledbyaLaplacianoraGammadistributionandtheDCTcoefficientsofthenoiseareGaussiandistributed.Then,MMSEestimatorsunderspeechpresenceuncertaintyarederived.Furthermore,theproperestimatorsofthespeechstatisticalparametersareproposed.ThespeechLaplacianfactorisestimatedbyanewdeci-sion-directedmethod.ThesimulationresultsshowthattheproposedalgorithmyieldslessresidualnoiseandbetterspeechqualitythantheGaussianbasedspeechenhancementalgorithmsproposedinrecentyears.
简介:风廓线雷达大气湍流回波信号的识别常受到间歇性杂波的严重干扰。为了能够有效抑制并消除该类杂波,本文提出了基于线性时-频变换的间歇性杂波消除抑制算法。该算法采用具有线性时-频分布特性的短时傅里叶变换,获取雷达回波中杂波成分的调制频率参数,然后估计出相应分数阶傅里叶变换(FRFT)的最优阶数,最后再利用FRFT检测出杂波并对其抑制,通过信号重构回到时域,从而获取有效回波信号频谱。本文提出的方法与常规的FRFT最优阶数扫描法相比,将杂波检测二维搜索问题简化为一维峰值搜索问题,计算量大大降低。理论及计算机仿真表明,该方法效果明显,具有良好的工程应用前景。