简介：Anoveltextindependentspeakeridentificationsystemisproposed.Intheproposedsystem,the12-orderperceptuallinearpredictivecepstrumandtheirdeltacoefficientsinthespanoffiveframesareextractedfromthesegmentedspeechbasedonthemethodofpitchsynchronousanalysis.TheFisherratiosoftheoriginalcoefficientsthenbecalculated,andthecoefficientswhoseFisherratiosarebiggerareselectedtoformthe13-dimensionalfeaturevectorsofspeaker.TheGaussianmixturemodelisusedtomodelthespeakers.TheexperimentalresultsshowthattheidentificationaccuracyoftheproposedsystemisobviouslybetterthanthatofthesystemsbasedonotherconventionalcoefficientslikethelinearpredictivecepstralcoefficientsandtheMel-frequencycepstralcoefficients.

简介：