用于自动语音识别系统的切换语音功率谱估计算法

被引:4
作者
刘金刚
周翊
马永保
刘宏清
机构
[1] 重庆邮电大学通信与信息工程学院
关键词
自动语音识别系统; 鲁棒性; 最小均方误差; 语音存在概率; 功率谱估计; 维纳滤波器;
D O I
暂无
中图分类号
TN912.34 [语音识别与设备];
学科分类号
0711 ;
摘要
针对语音识别系统在噪声环境下不能保持很好鲁棒性的问题,提出了一种切换语音功率谱估计算法。该算法假设语音的幅度谱服从Chi分布,提出了一种改进的基于最小均方误差(MMSE)的语音功率谱估计算法。然后,结合语音存在的概率(SPP),推导出改进的基于语音存在概率的MMSE估计器。接下来,将改进的MSME估计器与传统的维纳滤波器结合。在噪声干扰比较大时,使用改进的MMSE估计器来估计纯净语音的功率谱,当噪声干扰较小时,改用传统的维纳滤波器以减少计算量,最终得到用于识别系统的切换语音功率谱估计算法。实验结果表明,所提算法相比传统的瑞利分布下的MMSE估计器在各种噪声的情况下识别率平均提高在8个百分点左右,在去除噪声干扰、提高识别系统鲁棒性的同时,减小了语音识别系统的功耗。
引用
收藏
页码:3369 / 3373+3384 +3384
页数:6
相关论文
共 21 条
[1]  
Speech enhancement using a minimum mean-square error log-spectral amplitude estimator. Ephraim Y, Malah D. IEEE Transactions on Acoustics Speech and Signal Processing . 1985
[2]  
Optimal speech enhancement under signal presence uncertainty using log-spectral amplitude estimator. Cohen, Israel. IEEE Signal Processing Letters . 2002
[3]  
Generalized Gamma modeling of speech and its estimation for speech enhancement. Dat T H,Takeda K,Itakura F. International Conference on Acoustics, Speech and Signal Processing . 2005
[4]  
A novel a priori SNR estimation approach based on selective cepstro-temporal smoothing. BREITHAUPT C,GERKMANN T,MARTIN R. Proceedings of the 2008 IEEE International Conference on Acoustics Speech and Signal Processing . 2008
[5]  
Minimum mean-squared error estimation of mel-frequency cepstral coefficients using a novel distortion model. Indrebo K.M,Povinelli R.J,Johnson M.T. IEEE Transactions on Audio, Speech and Language Processing . 2008
[6]  
A study of complexity and quality of speech waveformcoders. Tribolet J,Noll P,McDermott B, et al. IEEE International Conference on Acoustics, Speech, Signal Processing . 1978
[7]  
Speech Enhancement, Theory and Practice. LOIZOU P C. CRCPress . 2007
[8]  
Minimum mean-square error estimation of Mel-frequency cepstral features theoretically consistent approach. JENSEN J,TAN Z H. IEEE/ACM Transactions on Audio,Speech and Language Processing . 2015
[9]  
Minimum Mean-Square Error Estimation of Discrete Fourier Coefficients With Generalized Gamma Priors. J. S. Erkelens,R. C. Hendriks,R. . Heusdens,J. . Jensen. IEEE Transactions on Audio, Speech, and Language Processing . 2007
[10]  
MMSE speech enhancement under speech presence uncertainty assuming(generalized)Gamma speech priors throughout. FODOR B,FINGSCHEIDT T. Proceedings of the 2012 IEEE International Conference on Acoustics Speech and Signal Processing . 2012