高阶谱分析在抗噪语音编码中的应用  

Application of Spectrum with Higher Order to Noise Robust in Speech Coding

作　　者：姚文冰[1] 姚天任[1] 韩涛[1] 

机构地区：[1]华中科技大学电子与信息工程系

出　　处：《华中科技大学学报（自然科学版）》2001年第9期60-62,共3页Journal of Huazhong University of Science and Technology(Natural Science Edition)

摘　　要：将高阶谱分析技术应用于语音编码中以提高其抗噪声的性能 .给出了两种利用高阶累积量提取语音参数的语音编解方案 :a.采用高阶累积量最小二乘直接估计法提取语音参数进行编码 ;b .采用高阶累积量SVD TLS估计法提取语音参数进行编码 .研究结果表明 ,这两种方案在噪声环境中工作时 ,具有极好的抗高斯有色噪声和对称分布噪声的能力 ,总的抗噪性能明显优于传统的LPC声码器 .The technology of analyzing spectrum with higher order is applied to speech coding to suppress noise. Two HOS based on speech coding proposals are presented: a. Extracting speech features with HOS based on least square estimation method. b. Extracting speech features with HOS based on Singular Value Decomposition Total Least Square estimation method. Research results show that both systems have very strong abilities in suppressing Gaussian colorful noise and symmetrically distributing noise. The two systems have more noise robust than the conventional LPC coding proposals have.

关 键 词：语音编码 高阶谱分析 稳健语音特征 线性预测分析 抗噪声性能 语音参数 

分 类 号：TN912.31[电子电信—通信与信息系统]