简介：

简介：基于数字听觉的信号的批评乐队小浪变换嵌入算法的水位标在这份报纸被建议。为每个听觉的信号片断的掩盖的阀值被计算在上psychoacoustic基本模型。根据人的听觉的系统和批评乐队小浪变换的批评乐队的类似，水位标被嵌进低乐队并且数字小浪的中等带宽系数。嵌入的力量被掩盖的阀值适应地控制。嵌入的水位标信号是听不见的实验结果表演，和watermarked声音信号对象压缩那样的许多攻击有好坚韧性，噪音，重新采样，低通行证的过滤。

简介：Foraccuracyandrapidityofaudioeventdetectioninthemass-dataaudioprocessingtasks,agenericmethodofrapidlyrecognizingaudioeventbasedon2D-HaaracousticsuperfeaturevectorandAdaBoostisproposed.Firstly,itcombinescertainnumberofcontinuousaudioframestobean'acousticfeatureimage',secondly,usesAdaBoost.MHorfastRandomAdaBoostfeatureselectionalgorithmtoselecthighrepresentative2D-Haarpatterncombinationstoconstructsuperfeaturevectors;thirdly,analyzesthecommonalityanddifferencesbetweensubcategories,thenextractscommonfeaturesandreducesdifferentfeaturestoobtainagenericaudioeventtemplate,whichcansupporttheaccurateidentificationofmultiplesub-classesanddetectandlocatethespecificaudioeventfromtheaudiostreamaccurately.Experimentalresultsshowthattheuseof2D-Haaracousticfeaturesupervectorcanmakerecognitionaccuracy5%higherthanonesthatMFCC,PLP,LPCCandothertraditionalacousticfeaturesyielded,andcanmakethetrainingprocessing7-20timesfasterandtherecognitionprocessing5-10timesfaster,itcanevenachieveanaverageprecisionof93.38%,anaveragerecallof95.03%undertheoptimalparameterconfigurationfoundbygridmethod.Aboveall,itcanprovideanaccurateandfastmass-dataprocessingmethodforaudioeventdetection.