基于小波包分解和噪声分析的抗噪说话人识别特征参数

Robust Feature Based on Noise Analysis and Wavelet Packet for Speaker Recognition

为提高说话人识别中语音特征参数对噪声的鲁棒性,本文提出在对语音进行小波包分解基础上,分析噪声的特性,在不同子带内进行谱减并设立权重,提出了一种新的语音特征参数多层美尔倒谱系数。仿真实验表明,与MFCC特征参数相比,ML-MFCC在噪声环境下具有更好的抗噪性能和说话人识别率。

To improve the performance of speaker recognition in noise environment, a robust feature Multilayer Mel eepstrum coefficient（ML-MFCC） based on noise analysis and wavelet packet for speaker recognition is proposed. Experiments show that MLMFCC performs better than MFCC in noise environment.