摘要
提出一种基于最小均方误差估计维纳滤波器的设计方法与Matlab实现。通过使用莱文森-德宾算法求解维纳-霍夫方程(Yule-walker方程),得到滤波器系数进行维纳滤波。加载Matlab中的语音数据handel,人为地加入高斯白噪声,分别计算加入噪声后信号的自相关Rxx和加入噪声后信号和理想信号的互相关Rxd。在输出端将信号较为精确地重现出来,而噪声却受到最大抑制。实测数据的处理结果证明经过维纳滤波后语音信号的噪声减弱,信噪比提高,较好地改进了语音信号质量。
Based on minimum mean square error （MMSE） estimation, in this paper we present a Wiener filter design method and its Matlab implementation. By using Levinson-Durbin algorithm to solve Wiener-Hopf equations （Yule-Walker equations）, the filter coefficients will be available for Wiener filtering operation. After loading the voice data Handel in Matlab and artificially adding Gaussian white noise, the autocortelation Rxx with the noise signal added and the cross-correlafion Rxd with the noise signal and the ideal signal added will be respectively calculated. The signal will be more accurately reproduced at output terminal, but the noise is maximally suppressed. Result of measured data processing proves that the noise of speech signals weakens after Wiener filtering applied. It also shows that the signal-to-noise ratio and the quality of the voice signal have been well improved.
出处
《计算机应用与软件》
CSCD
2015年第1期153-156,共4页
Computer Applications and Software
基金
河南省重点科技攻关项目(132102310003)
