毕业论文 基于AR模型的音频差错隐藏算法及其改进.docVIP

毕业论文 ：基于AR模型的音频差错隐藏算法及其改进 中文摘要 随着网络的飞速发展，用户通过因特网对音频的需要已经越来越广泛，音频成为了人们交流和获取信息的一个重要手段。对音频信息进行发送、传输、接受的过程中会不可避免的出现一些错误，因此对音频信号差错进行处理的需求也越来越多。 本文主要实现了对发生错误的音频信号进行错误隐藏的处理。采用了基于AR模型的方法，用burg算法估计AR模型系数，然后采用直接的线性预测，实现了对音频错误的隐藏，此方法为外推法。或者在计算出模型系数后构造镜像激励，逆滤波得到预测信号，并对该信号通过滤波器组分子带进行后处理，得到最后结果。 上述的外推法和镜像激励法，后者的运算复杂度要优于前者。对错误点数为2048的序列，后者选择的AR模型阶数700即可达到前者2048阶的效果。运算量减少了很多。 关键词：信号重建，AR模型，burg算法，多采样，滤波器组 ABSTRACT Since network is developing at very fast speed, users’ need to get audio information from the internet is rising. And audio has become a very important method of communication and getting information. There must be some errors when the audio signal is being sent, transmit, and recieved. So there will be a great need for signal reconstruction arithmetic. This passage presents an AR model-based method of signal reconstruction. First ,we use burg’s method to estimate AR model coefficient ,then get the reconstruct signal via straight extrapolate. To reduce the computational complexity, we try to modify this method. After get the coefficient , we contrast an entended excitation concatenated with it’s time-reversed version, get the reconstruct signal via inverse filtering ,and then via multirate post-processing ,we can get the final signal . Considering the computational complexity, the last method is better than the first. to conceal a signal that has a 2048 error samples , when using the last method we only need an AR model with order 700,to get the same effect with using the first method with order 2048. Keywords：’s method, multirate, filterbank 主要符号对照表 EC 音频差错隐藏 Error Concealment FEC 前向错误编码 Forward Error Coding UDP 用户数据报协议 User Datagram Protocol RTP 实时传输协议 Real-time Transport Protocol UEP 不对称误码保护 Unequal error protection AR 自回归 Autoregressive ARMA 自回归滑动平均 Autoregressive moving-average 目 录 第1章 引言 1 1.1 研究背景 1 1.1.1 音频差错隐藏的基本概念及其特点 1 1.1.2 音频差错隐藏的应用范围 2 1.1.3 研究动态 3 1.2 音频差错隐藏方法的分类 4 1.2.1 前向错误隐藏 5 1.2.2 交互式错误隐藏 6 1.2.3 后向错误隐藏 6 1.3 毕