音频信号截幅失真的检测与修复  

Detection and restoration of clipping distortion in audio signal

作　　者：刘春花[1] 付强[2] 杨家玮[1] 颜永红[2] 

机构地区：[1]西安电子科技大学通信工程学院,西安710071 [2]中国科学院声学研究所,北京100190

出　　处：《声学技术》2009年第4期498-502,共5页Technical Acoustics

基　　金：国家高技术研究发展计划(2006AA010102;2006AA01Z195);国家重点基础研究发展规划项目计划(2004CB318106);国家自然科学基金(10574140;60535030)资助

摘　　要：音频信号截幅失真的修复是信号修复工作的重要组成部分,目的是使得信号恢复完备性,以便于对其进行进一步研究。提出了一种音频信号截幅失真的检测与修复方法。首先通过研究截幅失真特性,描述了基于音频信号峰值大小而进行的检测算法,并创造性地将基于AR模型以及最小均方误差的LSAR算法应用于修复算法中。随后给出应用该算法后得到的实验结果,以及在截幅失真测试集上得到的信噪比及PESQ值改善的统计,由上述实验数据可以验证,该方法对不同采样率截幅失真的修复均有一定效果,达到了音频修复的要求,提高了音频质量。Clipping distortion in audio signal is an important part of signal restoration, whose aim is to make the audio integrative in order to do further research on it. This paper presents a method for detection and restoration of clipping distortions in audio signals. It begins with a description of the method for the detection of clipping distortions based on the audio＇s amplitude peak value, and creatively uses a least square AR （LSAR） repairing method based on the AR modeling and the theory of the minimum mean square error for the restoration of the clipping distortions. Subsequently, the results obtained by this method are shown, and the statistical results acquired by the method effecting on the clipping distortions test collection - the signal to noise ratio （SNR） and the value of the Perceptual evaluation of speech quality （PESQ） are presented. Obviously, the method is efficient when it processes on the clipping distortions with different sample rates, achieves the demands of audio restoration, and improves the audio quality.

关 键 词：数字截幅 模拟截幅 AR模型 LSAR 

分 类 号：TB556[理学—物理]