基于噪声环境下的MFCC特征提取

The MFCC Feature Extraction Based on The Noisy Situation

在语音识别系统中,如何在含有噪声的语音信号中提取其良好的特征参数是系统的关键问题之一。本文通过采用自适应滤波来滤除语音信号中的高斯白噪声,然后提取出能够反映人耳对语音感知特性的Mel频率倒谱系数(MFCC),并用其一阶差分参数加以改进,最后在实验中我们取得了较好的识别效果。

In the speech recognition system, how to pick up the characteristic parameter of the speech signal including the noise is one of the key problem. This paper removes the Gauss noise with the adaptive algorithm .Then the Mel Frequency Cepstral Coeffi- cients(MFCC) is extracted and with its difference parameter to improve. At last we get a better recognition effect in the experiment.