基于耳语频谱比较的话者识别方法  

Method of Voiceprint Recognition Based on Comparison of Whispered Chinese Speech Records' Spectrums

作　　者：董桂官[1] 沈勇[1] 

机构地区：[1]南京大学声学研究所近代声学教育部重点实验室,江苏南京210093

出　　处：《电声技术》2011年第4期51-52,66,共3页Audio Engineering

基　　金：国家自然科学基金项目(10774075)

摘　　要：通过分析由32人录制的143段汉语耳语语段的耳语频谱,发现同一人的耳语频谱之间的差异远小于不同人的耳语频谱之间的差异。基于此提出了一种通过比较耳语语段的耳语频谱差异,可用于话者识别的方法,并通过实验验证了该方法的可行性。143 whispered Chinese speech records spoken by 32 persons are analyzed in this paper. And the difference between one speaker＇s several spectrums of long-term average power spectral density is ｜ess than the difference between different speakers＇ ones. Then a method based on comparison of two whispered Chinese speech records＇ spectrums for voiceprint recognition is produced, and verification experiment shows it feasibility.

关 键 词：声纹识别 耳语频谱 汉语耳语 

分 类 号：TN912.3[电子电信—通信与信息系统]