Robust speech features representation based on computational auditory model

在线阅读 下载PDF 导出详情
摘要 Aspeechsignalprocessingandfeaturesextractingmethodbasedoncomputationalauditorymodelisproposed.Thecomputationalmodelisbasedonpsychological,physiologicalknowledgeanddigitalsignalprocessingmethods.Ineachstageofahearingperceptionsystem,thereisacorrespondingcomputationalmodeltosimulateitsfunction.Basedonthismodel,speechfeaturesareextracted.Ineachstage,thefeaturesindifferentkindsoflevelareextracted.Afurtherprocessingforprimaryauditoryspectrumbasedonlateralinhibitionisproposedtoextractmuchmorerobustspeechfeatures.Allthesefeaturescanberegardedastheinternalrepresentationsofspeechstimulationinhearingsystem.Therobustspeechrecognitionexperimentsareconductedtotesttherobustnessofthefeatures.Resultsshowthattherepresentationsbasedontheproposedcomputationalauditorymodelarerobustrepresentationsforspeechsignals.
机构地区 不详
出处 《声学学报:英文版》 2004年4期
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)