面向NDN和SVC编码的自适应流媒体传输与控制研究

The wide adoption of mobile smart devices and users’ interests in video creates an era where mobile video becomes the king. Present video distribution technologies face several challenges such as heterogeneity, dynamism and scalability. The research community tries to improve the scalability of video streaming services from three aspects: video codecs, adaptive video streaming and network infrastructure. However, IP network’s end-to-end communication paradigm is not suitable for large-scale video content distribution from the architectural perspective. This restricts the superiority of dynamic adaptive video streaming, and also increases the complexity of scheduling algorithms. Meanwhile, when AVC encoding is used, data segments of different bitrates are independent of each other, which makes it hard to efficiently utilize network caches, and also increases the risk of bitrate oscillation due to inaccurate bandwidth prediction caused by adaption algorithms. .Since named data networking(NDN), scalable video coding(SVC) and adaptive video streaming have quite consistent objectives and complementary technical solutions, this project tries to gracefully integrate the three technologies. The caching, transmission and congestion control policies in the NDN network for SVC encoded video segments, the adaptive scheduling algorithms on the client side, and the coordination mechanism between the client and the NDN network will be studied. A prototype system will be developed to verify the feasibility of the integration. This research is expected to improve the QoE of video streaming services and also to demonstrate the advantages of content-centric networking architectures in general.

智能移动设备的普及和用户对视频的亲睐造就了当今移动视频为王的时代。现有的视频分发技术面临着异构性、动态性和可扩展性等挑战。虽然研究界从视频编码、自适应流媒体技术和网络基础设施这三个维度来提升流媒体服务的可扩展性，但是IP网络端对端的通信模式难以从架构层面适应大规模视频内容分发，这制约了自适应流媒体技术的优势，也增加了算法复杂性。此外，采用AVC编码，不同码率的数据块之间相互独立，难以有效利用网络缓存，也加剧了自适应算法由于带宽预测失误而引起的比特率跳变的风险。.命名数据网络NDN、可扩展视频编码SVC和自适应流媒体技术三者目标一致、方法互补，本项目拟将三者有机融合，研究基于SVC编码的视频数据块在NDN网络中的缓存、传输和拥塞控制策略，终端的自适应调度策略，以及终端与NDN网络的协同机制，并研制一套原型系统，验证三者融合的可行性，提升视频流媒体服务的用户体验，佐证内容中心网络架构的优越性。

本项目围绕视频流媒体服务所面临的异构性、动态性和可扩展性三大挑战，从网络架构、客户端调度算法和可扩展视频编码这三个维度入手，将三者有机融合，突破面向NDN 和SVC 编码的自适应流媒体关键技术，设计与NDN 网络和SVC 编码相适配的客户端调度算法，提高视频流媒体的平均播放码率，降低码率抖动性、缓冲频率和缓冲时间，佐证以内容中心的网络架构的优越性。.项目的主要研究内容和成果如下：.首先，研究了NDN中的路由缓存和高效转发机制。为了解决路由缓存中的缓存隐藏问题，提出了Atomic Caching和On-the-fly Caching两种缓存替换策略，实验结果表明，即便路由缓存大小仅占整个FIB条目的0.25%，缓存命中率也高达75%。.其次，研究了NDN网络中SVC感知的缓存协同策略。提出了面向SVC的反转概率缓存策略SRPC与线型概率缓存策略SLPC，实验结果表明所提策略能有效地将基础层视频块推向用户边缘，将高层视频块缓存在内容源附近；提出了基于按需Off-Path缓存探索的多路径转发策略O2CEMF，实验结果表明能有效提高基于网络编码的NDN网络的缓存命中率。.最后，研究了基于SVC编码的视频流媒体客户端调度算法。通过对多部视频的SVC编码进行分析，发现了SVC编码视频块的层间相关性，提出了基于SVC编码层间相关性的视频流媒体调度策略，在此基础上，进一步提出了基于SVC编码层间相关性和空间相关性的360°视频流媒体调度策略，实验结果表明，利用SVC编码块的层间相关性和空间相关性可以有效缓解码率高估和码率低估的问题。.通过4年的研究，在国内外高水平期刊和会议共发表论文21篇，其中，CCF C类及以上论文13篇，申请国家发明专利2项。