听觉特性和语谱特性在说话人识别中的应用  

Speech Enhancement Method of Auditory and Speech Spectrum Characteristics Applied in Speaker Identification

作　　者：王红玲[1] 张飞云[1] 

机构地区：[1]许昌学院,河南许昌461000

出　　处：《电声技术》2009年第6期57-59,86,共4页Audio Engineering

基　　金：河南省教育厅自然科学研究资助计划项目(2008A510013)

摘　　要：大多数说话人识别系统当由实验室走向实际应用时,环境噪声的存在会造成其识别性能下降。为了提高噪声环境下说话人识别系统的识别性能,将基于听觉特性和语谱特性的语音增强技术作为预处理器,首先对语音信号进行降噪处理,提高输入信号的信噪比。实验证明,经过降噪处理的语音信号送入说话人识别系统,提高了系统的识别性能。Speaker identification is critical to the highly developed information society. In real environments, noise has a detrimental effect on speech signal, and degrades the performance of speaker identification system. Speech enhancement method based on the auditory and speech spectrum characteristics is used to reduce the white noise in the front-end. Experimental results show that the performance of speaker identification system is improved with the processed speech signal in the noise environments.

关 键 词：说话人识别 后验Bark信噪比 临界频带 临界频带带宽 

分 类 号：TN391.4[电子电信—物理电子学]