关于噪声估计主要有这么几类方法
(1)基于VAD的方法,假设噪声是平稳的,在噪声段估计功率谱,然后用在语音段;
(2)基于Minimum Statistic及其变种的方法,包括MS算法[1],MCRA算法[2],IMCRA算法[3]
(3)基于weighted noise estimation 的方法[4];
(4) 基于Data Driven的方法;
。。。还有很多其他种类方法
大多围绕着如何降低噪声谱估计方差,如何提高对非平稳信号的快速追踪能力等问题展开。
第二类被研究的最多,应用的也比较广泛,推荐几篇参考文献:
[1] Rainer Martin, "Noise Power Spectral Density Estimation based on Optimal smoothing and minimum Statistics" IEEE Trans 2001
[2] Israel Cohen, "Noise Estimation by Minima Controlled Recursive Averaging for Robust Speech Enhancement" IEEE SPL 2002
[3] Israel Cohen, "Noise Spectrum Estimation in Adverse Environments: Improved Minima Controlled Recursive Averaging" IEEE Trans 2003
[4] M. Kato, A. Sugiyama, "Noise suppression with high speech quality based on weighted noise estimation and MMSE STSA" IWAENC 2001 |