一种DCT域的语音增强改进算法  

An Improvement Algorithm of Speech Enhancement in DCT Domain

作　　者：高留洋[1] 桑振夏[1] 杨晓杰[1] 李文君[1] 

机构地区：[1]中国人民解放军63898部队

出　　处：《现代电子技术》2011年第23期59-62,共4页Modern Electronics Technique

基　　金：国家863计划重点项目(2008AA011001)

摘　　要：提出了一种DCT域的语音增强改进算法。该算法依据DCT域语音分量和噪声分量统计分布的不同,采用最大后验概率估计方法对语音分量进行精确估计,实现将语音信号从噪声环境中分离出来,从而有效地改善语音质量。在低信噪比条件下,该算法可以有效降噪,且增强效果较频域增强算法更好。An improvement algorithm of speech enhancement is proposed.Based on different distribution of voice component and noise component in DCT domain,this algorithm used maximum a posteriori probability（MAP） estimation method to estimate the speech component accurately,to separate the voice signal from noise environment,so as to effectively improve the voice quality.In the condition of low SNR,this algorithm is able to restrain noise effectively and perform better than the frequency domain algorithm.

关 键 词：语音增强 DCT 语音分量 噪声分量 最大后验概率估计 

分 类 号：TN912.35[电子电信—通信与信息系统]