﻿ 基于变窗长搜索的改进型噪声估计算法
 计算机系统应用  2018, Vol. 27 Issue (9): 124-129 PDF

Improved Noise Estimation Algorithm Based on Searching by Variable Window
HU An, GAO Yong
College of Electronics and Information Engineering, Sichuan University, Chengdu 610065, China
Abstract: The MCRA minimum recursive algorithm is accurate for the noise estimation, and the changes of noise power spectrum in a speech can be tracked accurately. However, if the noise power spectrum increases too much suddenly, the original algorithm needs a period of time to get the accurate noise, and in this adaptive period, it will leave strong residual noise and affect people’s hearing experience. This paper introduces a Voice Activity Detection (VAD) algorithm which uses the maximum log-likelihood ratio with energy-zero ratio, and an improved noise estimation algorithm on the basis of MCRA is obtained. Experimental simulation also proves that the improved algorithm is better than the original algorithm in noise estimation speed.
Key words: speech enhancement     maximum logarithmic likelihood ratio     energy-zero ratio     noise estimation     MCRA

1 引言

2 基于变窗长的MCRA改进算法

2.1 MCRA算法基本原理

 ${S_f}(k,l) = \sum\limits_{i = - N}^N {w(i)} |Y(k,l){|^2}$ (1)

 $S(k,l) = {\alpha _s}S(k,l - 1) + (1 - {\alpha _s}){S_f}(k,l)$ (2)

$l$ 不能被 $L$ ( $L$ 为定窗长)整除时:

 \left\{ {\begin{aligned} & {{S_{\min }}(k,l) = \min (S(k,l),{S_{\min }}(k,l - 1))} \\ & {{S_{\rm{tmp}}}(k,l) = \min (S(k,l),{S_{\rm {tmp}}}(k,l - 1))} \end{aligned}} \right. (3)

$l$ 能被 $L$ 整除时:

 \left\{ {\begin{aligned} & {{S_{\min }}(k,l) = \min ({S_{\rm {tmp}}}(k,l - 1),S(k,l))} \\ & {{S_{\rm {tmp}}}(k,l) = S(k,l)} \end{aligned}} \right. (4)

${S_{\min }}(k,l)$ 表示搜索窗内的最小功率谱值, ${S_{{\rm{tmp}}}}(k,l)$ 用来缓存最小功率谱值.

 ${S_r}(k,l) = \frac{{S(k,l)}}{{{S_{\min }}(k,l)}}$ (5)

 $p(k,l) = {\alpha _p}p(k,l - 1) + (1 - {\alpha _p})I(k,l)$ (6)

 $\begin{array}{l}{\lambda _d}(k,l + 1) = {\lambda _d}(k,l)p(k,l) + (1 - p(k,l))({\alpha _d}{\lambda _d}(k,l) \\\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;\;+ (1 - {\alpha _d})|Y(k,l){|^2})\end{array}$ (7)

 $Zcr(l) = \sum\limits_{n = 0}^{N - 1} {|{\rm{sgn}}[{y_l}(n)] - {\rm{sgn}}[{y_l}(n - 1)]|}$ (8)

 $Energy{\rm{(}}l{\rm{)}} = \sum\limits_{k = 0}^{N - 1} {y{\rm{(}}l,k{\rm{)}}}$ (9)

 $Log\_Energy{\rm{(}}l) = {\rm {lo}}{{\rm g}_{10}}{\rm{(}}1 + Energy{\rm{(}}l{\rm{)}}/a{\rm{)}}$ (10)

 $E{{cr(}}l{{) = Energy(l)/(Zcr(}}l{{) + }}b{{)}}$ (11)

 \begin{aligned}& \wedge (k,l) = \frac{{p(Y(k,l)|{H_1})}}{{p(Y(k,l)|{H_0})}} = \frac{1}{{\xi (k,l)}}\exp \left\{ {\frac{{\gamma (k,l)\xi (k,l)}}{{1 + \xi (k,l)}}} \right\}\end{aligned} (12)

$Y(k,l)$ 为第 $l$ 帧带噪话音信号的傅里叶变换的第 $k$ 个子带处的频谱系数. $p(Y(k,l)|{H_1})$ 是假设话音存在情况下的条件概率, $p(Y(k,l)|{H_0})$ 是假设话音不存在情况下的条件概率. $\xi (k,l)$ $\gamma (k,l)$ 分别为第 $l$ 帧话音信号的第 $k$ 个频点的的先验性噪比和后验信噪比. 两种信噪比分别可由DD判决法则[11]得到, 分别表示为以下公式:

 $\xi (k,l) = \frac{{{\lambda _x}(k,l)}}{{{\lambda _d}(k,l)}}$ (13)
 $\gamma (k,l) = \frac{{|Y(k,l){|^2}}}{{{\lambda _d}(k,l)}}$ (14)

 $LLR(l) = \log \wedge (l) = \frac{1}{N}\sum\limits_{k = 0}^{N - 1} {\log \wedge (k,l)}$ (15)

2.3 双窗法并行搜索

2.4 噪声估计更新

3 话音估计器

 $G(k,l) = {(\frac{{\xi (k,l)}}{{1 + \xi (k,l)}})^{p(k,l)}}G_{\min }^{1 - p(k,l)}$ (16)

 图 2 语音增强框图

 $\overset{\lower0.5em\hbox{$\smash{\scriptscriptstyle\frown}$}}{X} (k,l) = Y(k,l) * G(k,l)$ (17)

4 实验仿真

4.1 实验设计

 图 3 高斯白噪声环境下的算法对比

4.2 实验结果分析

5 总结

 图 4 超短波噪声环境下的算法对比

 [1] Martin R. Noise power spectral density estimation based on optimal smoothing and minimum statistics. IEEE Transactions on Speech and Audio Processing, 2001, 9(5): 504-512. DOI:10.1109/89.928915 [2] Cohen I, Berdugo B. Noise estimation by minima controlled recursive averaging for robust speech enhancement. IEEE Signal Processing Letters, 2002, 9(1): 12-15. DOI:10.1109/97.988717 [3] 郑成诗, 李晓东, 莫福源, 等. 一种噪声谱估计的快速跟踪算法. 中国声学学会2006年全国声学学术会议论文集. 厦门, 中国. 2006. 401–402. [4] Rangachari S, Loizou PC. A noise-estimation algorithm for highly non-stationary environments. Speech Communication, 2006, 48(2): 220-231. DOI:10.1016/j.specom.2005.08.005 [5] 贺慧芳, 马建芬. 基于语音帧检测和子带谱跟踪的噪声估计算法. 计算机工程, 2009, 35(20): 281-282. DOI:10.3969/j.issn.1000-3428.2009.20.099 [6] 熊晶. 语音增强中噪声估计的研究[硕士学位论文]. 兰州: 兰州交通大学, 2015. [7] 王旭贞. 汽车噪声下语音增强算法的研究与应用[硕士学位论文]. 杭州: 杭州电子科技大学, 2015. [8] Maragos P. Fractal aspects of speech signals: Dimension and interpolation. Proceedings of 1991 International Conference on Acoustics, Speech, and Signal Processing. Toronto, ON, Canada. 1991. 417–420. [doi: 10.1109/ICASSP.1991.150365] [9] Chang JH, Kim NS, Mitra SK. Voice activity detection based on multiple statistical models. IEEE Transactions on Signal Processing, 2006, 54(6): 1965-1976. DOI:10.1109/TSP.2006.874403 [10] 夏丙寅, 鲍长春. 适应噪声强度突变的噪声估计加速方法. 信号处理, 2013, 29(10): 1336-1345. DOI:10.3969/j.issn.1003-0530.2013.10.010 [11] Ephraim Y, Malah D. Speech enhancement using a minimum-mean square error short-time spectral amplitude estimator. IEEE Transactions on Acoustics, Speech, and Signal Processing, 1984, 32(6): 1109-1121. DOI:10.1109/TASSP.1984.1164453 [12] 张建伟, 陶亮, 周健, 等. 基于改进谱平滑策略的IMCRA算法及其语音增强. 计算机工程与应用, 2017, 53(1): 153-157. DOI:10.3778/j.issn.1002-8331.1504-0029 [13] Cohen I, Berdugo B. Speech enhancement for non-stationary noise environments. Signal Processing, 2001, 81(11): 2403-2418. DOI:10.1016/S0165-1684(01)00128-1