基于维纳过滤的IMCRA算法

The IMCRA algorithm based on wiener filtering

为了改善噪声谱估计算法对噪声的估计能力,提出基于维纳过滤的最小值控制递归平均(improved minimum control recursion average,IMCRA)算法。采用二级过滤技术,第一级利用参变维纳滤波算法过滤带噪语音得到语音频谱的估算值,计算其先验信噪比和后验信噪比,通过维纳滤波传递函数计算输出语音。第二级利用IMCRA算法对语音信号进行噪声估计,使用基本谱减法过滤噪声得到语音信号。对比实验结果表明,该算法提高了噪声的估计能力,同时语音信号的可懂度和清晰度也有所提高。

In order to improve the ability to estimate noise spectrum,the minima controlled recursive averaging（IMCRA）algorithm based on Wiener filtering isproposedi.By using two level filtration technologies,at the first stage the parametric Wiener filtering algorithm is used for noisy speech to estimate the speech spectrum value,then the spectrum of the speech is calculated.The estimated value of SNR and posteriori SNR are then used through the Wiener filter transfer function to calculate the output speech.At second stage,the IMCRA algorithm is used to estimate the noise of speech signal,and then the basic spectral subtraction speech enhancement is used to filter noise signal.Experimental results show that this algorithm can improve the ability to estimate the noise and the speech intelligibility and clarity at the same time.