短时非线性预测方法对汉语语音特性的研究  

Study on Chinese speech characteristics by short-term nonlinear prediction

作　　者：徐歆[1] 胡水清[1] 陶超[1] 杜功焕[1] 

机构地区：[1]南京大学声学研究所近代声学国家重点实验室,南京210093

出　　处：《应用声学》2003年第5期36-40,44,共6页Journal of Applied Acoustics

基　　金：国家自然科学基金(10074035和19834040)

摘　　要：本文应用Short改进了的短时非线性预测方法对正常语速的汉语音节和短语进行了分析研究,揭示了汉语语音中浊音和清音的短时非线性预测能力的差异,并且发现这种差异即使在强背景噪声下仍能用短时非线性预测方法加以辨别。这些为浊音和清音的切分提供了一种可能性手段。This paper studies the characteristics of Chinese speech using the short-term nonlinear prediction method improved by Short. It is found that the accuracy of the short-term predictions of voiced speech signal and unvoiced speech signal is distinctly different, even in a strong noise background. So the method offers a potential method to divide the voiced speech from the unvoiced.

关 键 词：短时非线性预测 汉语 语音 特性 短语 浊音 清音 差异 噪声 语音发声系统 

分 类 号：TN912.34[电子电信—通信与信息系统]