基于对偶原理的异构环境视频传输研究

With the growing integration of telecommunications networks, Internet of Things, and 5G, there is a tremendous demand for video services over heterogeneous wireless circumstance. As different types of user devices are accessing multimedia services over heterogeneous networks, it is important to build video transmission that is able to adapt to wireless network resources. This project aims at developing new theories and techniques for quality improvement in video communication over heterogeneous circumstance through exploitation the duality. To accomplish this goal, the following research tasks are conducted. 1) Developing multi-dimensional state distribution model of heterogeneous characteristics based on Copula distribution. This model bridges the diversity in heterogeneous circumstances with the use of the developed distribution functions. 2) Devising the rate-distortion and sum-capacity model for progressive source coding and channel coding in heterogeneous circumstances, based on the duality principle. Such a model draws up the guide line for near-capacity designs and maximizes the performance of transmission. 3) Investigating the duality between video rates and utilization by using the polymatroidal structure to build a transmission model that supports heterogeneous devices. Such a module also develops a novel mechanism that uses Max-Min fairness and distributed resource allocation for cooperative transmission. This thereby provides fair video transmission among multiple heterogeneous devices. 4) Establishing the delay model by studying the duality between video rates and utilization under transmission-delay conditions, based on joint dirty-paper coding and Fountain coding. It eliminates the interference among multiple streaming and provides the delay optimal-transmission policy in heterogeneous wireless circumstances. This project is expected to broaden the understanding of QoE-assured heterogeneous video communication. The proposed techniques will fill the gap between fundamental theories and practical heterogeneous transmission by providing the theoretical underpinning for systematic design and algorithmic implementation.

随着三网融合、物联网和第五代移动通信技术的发展，在异构环境中提供高质量稳健的视频流成为视频和通信网络领域研究的热点问题，由于终端种类日益增多，在无线视频传输中，提高多异构终端用户的视频传输性能成为新趋势，受到了国内外学术界和工业界的广泛关注。本项目拟从研究异构环境视频传输中的基础模型着手，通过挖掘对偶关系来探索异构环境下提供高质量可靠视频传输的新方法。主要研究内容包括：1）分析异构环境的特点，建立基于Copula分布的异构终端多维状态分布模型；2）针对多异构特征,通过对偶原理建立渐近式信源编码与信道编码之间的率失真和合容量模型；3）研究码率与终端效用之间的对偶，采用多拟阵类建立资源分配对偶问题的几何模型，从而优化多异构终端的视频公平传输；4）研究延时约束下的对偶，通过脏纸编码思想去除多信源间的干扰，并联合喷泉码建立延时优化模型，从而使多视频流在有限带宽下，多异构终端获得最优实时传输。

本项目致力于提高面向异构终端视频传输性能，项目执行期间取得了很大的进展，按照计划书分别在视频多维特征分析、视频处理中的对偶理论、基于对偶理论的传输架构和优化都取得了相应的研究成果。具体包括：（1）在视频多维特征分析方面，提出了基于Copula分布的多维特征联合分析方法，可以有效解决多维特征异构化造成的数学建模困难问题；（2）针对对偶理论，本项目研究了视频处理中存在的对偶关系，并依据信道容量与率失真之间的对偶性，建立了无反馈的分布式视频编码的结构（3）研究了码率与终端效用之间的对偶性，提出基于多拟阵类建立异构环境下资源分配对偶问题的几何模型，解决了异构环境下资源优化的效率低问题；（3）研究了时延约束下的对偶性，提出基于时限感知的码率分配算法，从而使多视频流在有限带宽下，多异构终端获得最优实时传输。.本项目执行期间，共发表和录用24篇著名国际期刊或会议论文，其中已录用2篇国际期刊论文，已发表15篇国际期刊论文，已发表7篇国际会议论文，并申请国家发明专利9项，转让3项国家发明专利，培养了多名博/硕士研究生，项目负责人及研究成果还获得2016年北京市科学技术奖二等奖（个人排名第5），2016年中国专利优秀奖（个人排名第3），2017年CCF科学技术奖技术发明一等奖（个人排名第3）。圆满完成了项目研究目标。