基于短时对数谱的MMSE语音增强算法研究

研究了单话筒采集条件下基于语音短时对数谱的最小均方误差(MMSE—LSA)估计的语音增强算法，给出了其算法分析的基本流程图。由于语音是时变的，因此，假设语音频谱分布为高斯分布，在此基础上讨论了MMSE—LSA算法的先验信噪比ξh的2种估计方法——最大似然估计方法和直接判决估计方法。试验证明此方法的语音增强效果较好，尤其在较低信噪比时效果更明显。

Study of speech enhancement based on MMSE short time log spectral amplitude estimator

In this paper,the authors study the speech enhancement algorithm based on minimun mean square error short time log spectral amplitude estimation(MMSE-LSA) under the single input condition,and present the basic flowchart of the algorithm analysis.Because the speech is non-stationary,and supposing the speech spectrum is the distribution of Gaussian,the authors discuss the estimation method of the parameter-priori SNR(signal-to-noise ratio)-maximum likelihood estimation approach and decision-directed estimation approach.Experiments show that the performance of the algorithm enhances the speech very well, especially in the conditions of low SNR.