基于MDCT的线性带宽扩展方法  

A MDCT based linear bandwidth extension method

作　　者：李思源 姜林 LI Siyuan;JIANG Lin(College of Information Engineering,East China University of Technology,Nanchang 330013,China;College of Computer and Information Engineering,Hunan University of Technology and Business,Changsha 410205,China)

机构地区：[1]东华理工大学信息工程学院,南昌330013 [2]湖南工商大学计算机与信息工程学院,长沙410205

出　　处：《智能计算机与应用》2020年第3期69-71,75,共4页Intelligent Computer and Applications

基　　金：国家自然科学基金(61762005);湖南省自然科学基金(2019JJ40144);湖南省教育厅科学研究项目(18A304)。

摘　　要：音频带宽扩展技术是音频编解码技术的重要组成部分,研究高频信号的编码方法,目的是大幅减少编码码率。针对传统频域率带宽扩展框架中直接高频信号在频率域编码导致高码率和高复杂度的问题,本文提出基于MDCT变换的带宽扩展编码方法,该方法基于源滤波器模型对高频信号进行分析合成。编码中对高频信号进行LPC参数提取,并将高频激励信号进行MDCT变换,提取MDCT子带能量系数,然后对LPC和MDCT子带能量进行矢量编码并传送至解码端。解码时将低频激励作为高频激励信号的精细结构,采用MDCT子带能量对精细结构进行调整,最后采用LPC合成得到高频重建信号。实验结果表明,本文方法相比经典的SBR方法,主观音质相当,编码码率下降了56.36%,整体时间复杂度下降明显。The audio bandwidth extension technology is an important part of the audio codec technology.It studies the encoding method of audio compression,which can greatly reduce the coding rate.For the traditional frequency domain rate bandwidth extension framework,the time-frequency transform of the original audio signal after passing the filter directly leads to the problem of coding bit rate higher.The linear encoder based on the linear prediction is used in the traditional source filter bandwidth method.In combination with the MDCT transform,the residual signal obtained by the LPC analysis is firstly used as the fine structure of the high-frequency signal,and the fine structure is subjected to MDCT transform to obtain the fine-domain frequency domain signal,and then the frequency domain of the fine structure is achieved.The signal is encoded to obtain the final reconstructed high frequency signal.The experimental results show that the encoded sound quality is equivalent to the SBR algorithm,the coding rate is reduced by 56.36%,and the complexity is also reduced.

关 键 词：带宽扩展 音频压缩 线性预测 MDCT变换 

分 类 号：TP37[自动化与计算机技术—计算机系统结构]