提高MFCC鲁棒性的方法

Approach to improve robust performance of Mel frequency coefficient cepstral

在噪声环境中,基于笔记本电脑录音的情况下,采用特征参数加窗的方法,以提高系统的噪声鲁棒性。在Matlab环境下,建立了基于高斯混合模型(GMM)的说话人辨认系统,并进行实验。通过对多种窗口的正识率比较,发现对美尔倒谱(MFCC)高阶参数的加窗提升,可以改善系统的鲁棒性。实验结果表明,采用加窗后的系统识别率得到了明显改善。

An approach of the cepstal liftering window is used to improve robust of the speaker recognition system under the Notebook- PC condition in noise surrounding. A speaker identification system based on Gaussian mixture model （GMM） is applied to an experiment for the cepstral coefficient in Matlab. With compared different liftering windows in the result of right recognition rate, we found that the high cepstral term of MFCC make better robust performance in speaker recognition. The result shows that the performance of the system robust with liftering window is further improved.