机构地区: 华南理工大学电子与信息学院
出 处: 《成都信息工程学院学报》 2004年第4期541-544,共4页
摘 要: 详细介绍了说话人确认的基本原理和主要技术方法,并通过对传统高斯混合模型(GMM)匹配方法的分析,指出了其运算量大、自适应性差的缺点,并在此基础上提出了解决方法-变阈值判决准则(DTSV)。 The basic principle and main techniques of the speaker recognition is discussed. The traditional GMM is analyzed and its defects such as heavy computation and poor self-adaptation are pointed out. Finally a solution-DTSV to those problems is presented.