摘要： 为了提高电话语音识别系统在环境和信道等噪声下的鲁棒性，文章针对电话语音MFCC特征的低阶系数最易受到污染的特点，通过舍弃最易受污染的低阶MFCC系数，并结合RASTA滤波来去除卷积信道噪声，使系统的识别性能和抗噪性能有了明显的改善。

关键词: 语音识别, 带通滤波器, 高阶MFCC, 鲁棒性

Abstract: The paper proposes a method in order to suppress the channel noise and improve the robustness of telephone speech recognition. In the method , RelAtive SepeTrAl(RASTA) filter is used to reduce the convolution channel noise and the low_order mel-frequency cepstral coefficients(MFCC) is omitted because it's usually easy contaminated. Experiments show that the method can greatly improve system's recognition performance.

中图分类号: 

• TN912.3