资源描述:
《基于段级特征主成分研究说话人识别算法》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、基于段级特征主成分研究说话人识别算法 文章编号:10019081(2013)07193503doi:10.11772/j.issn.10019081.2013.07.1935摘要:为了提高说话人识别(SR)系统的运算速度,增强其鲁棒性,以现有的帧级语音特征为基础,提出了一种基于段级特征主成分分析的说话人识别算法。该算法在训练和识别阶段以段级特征代替帧级特征,然后用主成分分析方法对段级特征进行降维、去相关。实验结果表明,该算法的系统训练时间、测试时间分别为基线系统的47.8%、40.0%,同时识别率略有提高,
2、抑制了噪声对说话人识别系统的影响。该结果验证了基于段级特征主成分分析的说话人识别算法在识别率有所提高的情况下取得了较快的识别速度,同时在不同噪声环境下的不同信噪比情况下均可以提高系统识别率。关键词:说话人识别;非线性分段;主成分分析;说话人识别系统中图分类号:TP18文献标志码:A英文标题Speakerrecognitionmethodbasedonutterancelevelprincipalcomponentanalysis英文作者名5CHUWen1,2*,LIYinguo2,XUYang2,MENGXia
3、ngtao1,2英文地址(1.CollegeofComputerScienceandTechnology,ChongqingUniversityofPostsandTelecommunications,Chongqing400065,China;2.ResearchCenterofAutomotiveElectronicsandEmbeddedSystemEngineering,ChongqingUniversityofPostsandTelecommunications,Chongqing400065,China
4、英文摘要)Abstract:ToimprovethecalculationspeedandrobustnessoftheSpeakerRecognition(SR)system,theauthorsproposedaspeakerrecognitionalgorithmmethodbasedonutterancelevelPrincipalComponentAnalysis(PCA),whichwasderivedfromtheframelevelfeatures.Insteadofframelevelfeatures,this
5、algorithmusedtheutterancelevelfeaturesinbothtrainingandrecognition.Whatsmore,thePCAmethodwasalsousedfordimensionreductionandredundancyremoving.Theexperimentalresults5showthatthisalgorithmnotonlygetsalittlehigherrecognitionrate,butalsosuppressestheeffectofthenoiseonsp
6、eakerrecognitionsystem.ItverifiesthatthealgorithmbasedonutterancelevelfeaturesPCAcangetfasterrecognitionspeedandhighersystemrecognitionrate,anditenhancessystemrecognitionrateindifferentnoiseenvironmentsunderdifferentSignaltoNoiseRatio(SNR)conditions.Toimprovethecalc
7、ulationspeedandrobustnessoftheSpeakerRecognition(SR)system,theauthorsproposedamethodbasedonutterancelevelPrincipalComponentAnalysis(PCA)ofspeakerrecognitionalgorithm,whichisderivedfromtheframelevelfeatures.Insteadofframelevelfeatures,thisalgorithmusedtheutterancelevel
8、featuresbothintrainingandrecognition.What’smorethePCAmethodisalsousedfordimensionreductionandredundancyremoving.Theexperimen