摘要
Aspeechsignalprocessingandfeaturesextractingmethodbasedoncomputationalauditorymodelisproposed.Thecomputationalmodelisbasedonpsychological,physiologicalknowledgeanddigitalsignalprocessingmethods.Ineachstageofahearingperceptionsystem,thereisacorrespondingcomputationalmodeltosimulateitsfunction.Basedonthismodel,speechfeaturesareextracted.Ineachstage,thefeaturesindifferentkindsoflevelareextracted.Afurtherprocessingforprimaryauditoryspectrumbasedonlateralinhibitionisproposedtoextractmuchmorerobustspeechfeatures.Allthesefeaturescanberegardedastheinternalrepresentationsofspeechstimulationinhearingsystem.Therobustspeechrecognitionexperimentsareconductedtotesttherobustnessofthefeatures.Resultsshowthattherepresentationsbasedontheproposedcomputationalauditorymodelarerobustrepresentationsforspeechsignals.
出版日期
2004年04月14日(中国期刊网平台首次上网日期,不代表论文的发表时间)