摘要
MCRA(Minima-Controlled Recursive Averaging)方法是经典的噪声估计算法,然而在语音段MCRA方法存在不能对噪声功率谱进行有效更新的问题.针对这一问题,本文利用广义自回归条件异方差(Generalized Autoregressive Conditional Heteroskedasticity,GARCH)模型在时频域对噪声信号建模,在MCRA算法原理的基础上,提出了基于最小控制GARCH模型的噪声估计算法,实验结果表明,本文所提的噪声估计算法能够更为准确估计噪声功率谱,将该算法应用到语音增强中能够获得到较好的语音增强效果.
Considering the problem that the typical M CRA( M inima-Controlled Recursive Averaging) noise estimate algorithm fails to update the pow er spectrum of noise effectively w hen the speech is present,so this paper proposes a noise estimate algorithm based on minima controlled GARCH model. The noise signal is modeled as a GARCH process in timefrequency domain and then the proposed noise estimate algorithm is achieved combined w ith the basis of the framew ork of M CRA method. Experimental and testing results indicate that the proposed algorithm can estimate the spectrum of noise more accurately compared w ith the reference methods. When the proposed algorithm is applied into speech enhancement,a better performance can be achieved as w ell.
引文
[1]Hendriks R C,Heusdens R,Jensen J.MMSE based noise PSD tracking w ith low complexity[A].2010 International Conference on Acoustics,Speech,and Signal Processing[C].Dallas,Texas,USA:IEEE Press,2010.4266-4269.
[2]张东方,蒋建中,张连海.一种改进型IMCRA非平稳噪声估计算法[J].计算机工程,2012,38(13):270-272.Zhang Dong-fang,Jiang Jian-zhong,Zhang Lian-hai.Improved IM CRA non-stationary noise estimation algorithm[J].Computing Engineering,2012,38(13):270-272
[3]Martin R.Spectral subtraction based on minimum statistics[J].European Signal Processing Conference,1994,1182-1185.
[4]Cohen I,Berdugo B.Noise estimation by minima controlled recursive averaging for robust speech enhancement[J].Signal Processing Letters,IEEE,2002,9(1):12-15.
[5]Cohen I.Modeling speech signals in the time-frequency domain using GARCH model[J].European Signal Processing,2004,84(12):2453-2459.
[6]Cohen I,Speech spectral modeling and enhancement based on generalized autoregressive conditional hetero-skedasticity models[J].European Signal Processing,2006,86(4):698-709.
[7]Loizou P.Speech Enhancement:Theory and Practice[M].Boca Raton,FL,USA:CRC Press,2007.435-439.
[8]Kamath S,Loizou P.A multi-band spectral subtraction method for enhancing speech corrupted by colored noise[A].IEEE International Conference on Acoustics Speech and Signal Processing[C].Orlando,Florida,USA:IEEE,2002.4164-4164.
[9]周旋.基于统计模型和经验模式分解的宽带语音增强技术研究[D].北京:北京工业大学,2012.