Abstract: The trade-off among amount of noise reduction, the level of musical residual noise and the speech distortion in high noise environment has to be solved. This study focuses on noise estimation to boost the performance of speech enhancement algorithms by combining statistical estimators of the spectral magnitude of the speech and noise. The noise spectral magnitude estimator is derived from the speech magnitude estimator, with which a modified a priori SNR estimation is given. The simulated experiments indicate that significant improvement could be achieved. For the speech enhancement, the algorithm obtains an obvious improvement in reducing speech distortion, with little musical noise retained.

Key words: high noise, noise estimation, a priori Signal Noise Ratio（SNR）, speech present probability, speech enhancement

摘要： 针对强噪声环境下语音增强中噪声估计和先验信噪比估计算法导致的语音失真和音乐噪声的问题，利用语音和噪声的统计模型的对称性得到一种噪声幅度的估计值为参考，提出了一种噪声估计算法，改进了先验信噪比估计算法，形成了一种新的增强算法，适用于强噪声环境下的语音增强。由仿真实验给出的客观评分看出，在0 dB乃至－5 dB条件下，给出信噪比估计算法能够有效减小信号失真，基本上没有残留音乐噪声。

关键词: 强噪声, 噪声估计, 先验信噪比, 语音存在概率, 语音增强