|dc.description.abstract||With the advancement of video compression technology and wide deployment of wired/wireless networks, there is an increasing demand of multiuser video communication services. A multiuser video transmission system should consider not only the reconstructed video quality in the individual-user level but also the service objectives among all users on the network level. There are many design challenges to support multiuser video communication services, such as fading channels, limited radio resources of wireless networks, heterogeneity of video content complexity, delay and decoding dependency constraints of video bitstreams, and mixed integer optimization. To overcome these challenges, a general strategy is to dynamically allocate resources according to the changing environments and requirements, so as to improve the overall system performance and ensure quality of service (QoS) for each user.
In this dissertation, we address the aforementioned design challenges from a resource-allocation point of view and two aspects of system and algorithm designs, namely, a cross-layer design that jointly optimizes resource utilization from physical layer to application layer, and multiuser diversity that explores the source and channel heterogeneity among different users. We also address the impacts on systems caused by dynamic environment along time domain and consider the time-heterogeneity of video sources and time-varying characteristics of channel conditions. To achieve the desired service objectives, a general resource allocation framework is formulated in terms of constrained optimization problems to dynamically allocate resources and control the quality of multiple video bitstreams.
Based on the design methodology of multiuser cross-layer optimization, we propose several systems to efficiently transmit multiple video streams, encoded by current and emerging video codecs, over major types of wireless networks such as 3G cellular system, Wireless Local Area Network, 4G cellular system, and future Wireless Metropolitan Area Networks. Owing to the integer nature of some system parameters, the formulated optimization problems are often integer or mixed integer programming problem and involve high computation to search the optimal solutions. Fast algorithms are proposed to provide real-time services. We demonstrate the advantages of dynamic and joint resource allocation for multiple video sources compared to static strategy. We also show the improvement of exploring diversity on frequency, time, and transmission path, and the benefits from multiuser cross-layer optimization.||en_US