摘要
本文提出了1种基于小波域计算Teager能量自相关系数的语音端点检测算法。对语音信号进行3层小波分解,取其近似部分计算Teager能量及自相关系数,并以自相关系数的平均值作为端点检测门限。与单纯的Teager能量及传统的几种端点检测算法相比,此算法不仅更能表明语音段与静音段的特征,同时抗噪性能更优,对几种噪声类型语音均有一定的稳健性。
This paper proposes an end point detection algorithm based on calculating the auto - correlation coefficient of Teager energy on the wavelet domain. After Three layers of the speech signal wavelet decomposition, the auto - correlation coefficient of the Teager energy of the speech signal is calculated; and the average of the auto - correlation coefficient is enacted as the threshold. Compare to the simple Teager energy and the number of traditional end point detection algorithms, the auto - correlation coefficient of the Teager energy possesses good performance in distinguishing the voiced segments and unvoiced ones, and can adapt to circumstance that the speech signal has different SNR or several different types of noise.
出处
《煤炭技术》
CAS
北大核心
2010年第1期178-181,共4页
Coal Technology
