基于核Fisher判别和加权码书映射的音频信号削波修复方法  

Clipping Restoration of Audio Signals Based on Kernel Fisher Discriminant and Weighted Codebook Mapping

作　　者：邓峰[1] 鲍长春[1] 鲍枫[1] 

机构地区：[1]北京工业大学电子信息与控制工程学院,北京100124

出　　处：《数据采集与处理》2014年第2期211-221,共11页Journal of Data Acquisition and Processing

基　　金：北京市教育委员会科技发展计划(KZ201110005005)重点资助项目;国家自然科学基金(61072089)资助项目

摘　　要：基于核Fisher判别(Kernel fisher discriminant,KFD)和加权码书映射(Weighted codebook mapping,WCBM),提出了一种MDCT(Modified discrete cosine transform)域的音频信号削波修复方法。首先根据音频信号的MDCT系数提取子带包络等四种削波特征参数;其次,利用这些特征参数训练检测音频信号出现削波的核Fisher分类器;最后,利用子带包络的WCBM来修复音频信号的削波。测试结果表明,本文所提方法能有效修复音频信号的削波,其性能优于现有的几种削波修复方法。A new clipping restoration method of audio signals is proposed based on kernel Fisher diseriminant （KFD） and weighted codebook mapping （WCBM） in modified discrete cosine transform （MDCT） domain. Firstly, the four clipping features are extracted according to the MDCT coefficients of audio signals. Secondly, the clipping feature parameters are used to train the optimal kernel Fisher classifier which is employed to detect the clipping. Finally, the WCBM of sub-band envelop is adopted to restore the clipping of audio signals. The test results indicate that the proposed method can effectively remove the clipping distortion of audio signals and obviously outperforms the existing clipping restoration methods.

关 键 词：核FISHER判别 加权码书映射 修正离散余弦变换 削波修复 削波检测 

分 类 号：TN912.3[电子电信—通信与信息系统]