摘要
针对无线环境中基于信息包信道分集技术的多重描述编码进行了研究.在比较了多重描述编码(multipledescriptioncoding,简称MDC)和分层编码(layercoding,简称LC)的性能特点并总结了它们各自的优点与不足之后,提出一种能有效结合两者长处的新的编码方法——自动恢复的多重描述编码(auto-resilientmultipledescriptioncoding,简称ARMDC)方案.ARMDC既能够像通常的MDC一样使每个成功到达的数据包都提供有效信息,也能像LC一样优先传送某一重要的描述,保证基本的视频质量,且解码端接收到的ARMDC码流可以进行周期性的自动恢复,无须等待源端接收到反馈信息后再作出反应.在仿真的3G无线信道上分别传送ARMDC,MDR(multipledescriptionwithrestart)以及不使用LC和MDC的码流,比较其各自的错误恢复性能.实验结果显示,使用ARMDC的视频流在序列解码后有着更好的R-D性能.同时,由于视频序列相邻的帧之间质量较平滑,也获得了更好的主观性能.此研究成果可以应用于无线媒体流业务与无线会话业务中的视频传输系统,对于研究基于易出错时变信道的多媒体传输的同行具有一定的参考价值.
In this paper, a source coding technique exploiting packet path diversity is presented, which combines the individual benefits of MDC （multiple description coding） and LC （layer coding） for video transmission over wireless networks, Performance comparisons between LC and MDC are discussed at the beginning of this paper. Their advantages and disadvantages are concluded. An improved coding technique named auto-resilient multiple description coding （ARMDC） is then shown for its ability to conjoin the merits and avoid the shortcomings of the above two coding schemes. Under the critical transmission conditions of error -prone wireless channels, ARMDC can make all the received packets useful, not just those consecutive from the first packet. At the same time, it can also process the feature of LC, creating a ＂base-layer＂ bit-stream to ensure the basic quality of video. Transmission policy of ARMDC varies with transmission environment and application limits t o gain fine video quality. 3G mobile channels are simulated and then streams create by ARMDC, MDR （ multiple description with restart） scheme and coding scheme without any LC or MDC are sent over the channels. Experiment results show that the decoded video sequences of ARMDC streams have better performance in R-D curve comparision than the other MDC scheme and also gain best personality effect because of the smooth sequence quality. This out -coming is well adapted to streaming and conversational applications in wireless environment.
出处
《软件学报》
EI
CSCD
北大核心
2006年第3期371-378,共8页
Journal of Software
基金
国家自然科学基金
国家高技术研究发展计划(863)~~
