基于实值离散Gabor变换维纳滤波语音增强方法.doc

基于实值离散Gabor变换维纳滤波语音增强方法.doc基于实值离散Gabor变换维纳滤波语音增强
方法
摘要：提出了一个新的基于实值离散Gabor变换的维纳滤 波语音增强方法，采用高斯窗作为综合窗，利用已有的快速 实值离散Gabor变换将语音变换到时频域，然后在联合时频 域，采用维纳滤波进行纯净语音的最小均方误差下的最优估 计，先验信噪比采用“直接判决”算法，在得到语音增强信 号的估计分量后,利用实值离散Gabor逆变换将其还原输出。 实验结果表明，在分段信噪比和语音质量评价方面均与经典 的维纳滤波方法相比均有提高。
关键词：实值离散Gabor变换；噪声估计；维纳滤波； 语音增强
中图分类号：TP391文献标识码：A文章编号： 1009-3044(2012)17-4231-04
Speech Enhacement Approach Based on Real-value Discrete Gabor Transform and Spectral Subtraction
ZHANG Man, ZHOU Jian
(MOE Key Laborato【y of Intelligence Computing and Signal Processing, Anhui University, Hefei 230039, China)
Abstract： In this paper, we porpose a novel speeeh enhancement method based on Real-value Discrete Gabor Transform, noisy speech is transfonned to the joint time—frequency by fast Real-value Discrete Gabor Transform in which the Gaussin is used as the transform kernel function,0ptimal pure speech signal estimation under the minimum mean square error based on Wiener filter , a priori SNR got by deci? sion-directed algor it hm, the clean speech is got by inverse tr ansform by Real-value Discrete Gabor Transform. Exprimental resuIts show that this method enhances the segSNR and improves the speech quality compare other classical alogorithm of wiener filter.
Key words： mReal-value Discrete Gabor Transform； noise estimate； Wiener filter； speech enhacement 现实世界中，语音信号很难避免受到各种噪声源的污 染，语音增强的一个主要目的就是从带噪信号中尽可能的恢 复出纯净语音信号。语音增强技术在语音识别、语音编码和 人机语音交互等