基于MMSE-LSA语音增强算法在非平稳环境下的研究与实现

Study of speech enhancement algorithm based on MMSE-LSA under non-stationary environments

讨论了非平稳环境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。众所周知,语音信号为时变信号,在假设语音频谱分布为高斯分布的前提下,实验的工作重点是将MMSE-LSA算法与其它语音增强算法(以谱相减的语音增强为例)比较。实验结果表明:该MMSE-LSA算法的语音增强效果很好,特别是在信噪比低时的非平稳环境下效果更为明显。

The authors studied speech enhancement algorithm based on MMSE-LSA （minimum mean square error short time log spectral amplitude estimation） under stationary and non-stationary environments. Because the speech is non-stationary, and supposing the speech spectrum is the distribution of Gaussian, the method of Speech Enhancement algorithm different from the others is discussed. Experiments show that the performance of the algorithm enhances the speech very well, especially in the conditions of low SNR （signal-to-noise ratio） under stationary and non-stationary environments.