基于随机共振的说话人识别方法  

Speaker Recognition Method Based on Stochastic Resonance

作　　者：潘平[1] 杨平[1] 何朝霞[1] 

机构地区：[1]贵州大学计算机科学与信息学院,贵阳550023

出　　处：《电信科学》2010年第S2期74-77,共4页Telecommunications Science

基　　金：国家科技计划基金资助项目(No.2008RR0003);贵州省国际科技合作计划基金资助项目(No.[2009]700109;No.[2009]700125)

摘　　要：说话人识别研究,长期以来是基于对语料训练,构建语音库来识别说话人,其典型的方法是GMM、HMM等。由于语音信号的时变、混沌特性,说话人独一无二的特征一直是说话人识别的主要问题。论文以生理声学和非线性随机共振理论为基础,提出了一种基于随机共振的说话人识别方法,为说话人识别技术与方法研究提供了一条新的研究方向。The typical methods of speaker recognition,which recognize speaker by means of construction of speech database based on the training corpus for a long time,are GMM,HMM and etc.Owing to time-varying and chaotic characteristic of voice signal,the unique feature of speaker is the main issue of speaker recognition.In view of the physiological acoustics and the theory of nonlinear stochastic resonance,a method of speaker recognition due to stochastic resonance is proposed,which provides a new direction of technology and methods of speaker recognition.

关 键 词：说话人识别 随机共振 非线性 

分 类 号：TN912.34[电子电信—通信与信息系统]