Study On Key Techniques Of Media Streaming Over The Internet

With the rapid development of the Internet and multimedia compression technologies, itis possible to provide streaming media applications over the Internet. However, due to thebest-effort property of the Internet and the heterogeneity of networks and terminals, it isnecessary that the streaming media system must provide some capabilities of robustness andadaptiveness. Since scalability is one of the most important functions pursued by the nextgeneration multimedia coding standards, the research of this paper puts its emphasis mainlyon several key problems about the robust and adaptive streaming of scalable multimedia overthe Internet. Moreover, this paper also studies the implementation of a peer-to-peer streamingmedia system which is based on the Chinese AVS standard, and two network adaptiveframe-drop algorithms adopted in this system. The main research contents and contributionsof this paper are listed as follows:Firstly, in high bit rate conditions, if single unequal error protection framework is stillused to protect and transmit a scalable multimedia bitstream, the available channel bandwidthresources can not be fully utilized, and the receiver can not obtain better picture qualities withthe increased bandwidth. To address this problem, this paper proposes two grouped unequalerror protection schemes: grouped unequal error protection by splitting with packetinterleaving, and grouped unequal error protection by interleaving with packet interleaving.Experimental results show that both the two schemes can achieve near optimal averageperformance in various source and channel conditions, while the latter can still maintain theprogressive transmission performance of scalable source coding in a very good extend.Secondly, this paper studies the adaptiveness problems of multicasting unequal errorprotected scalable multimedia in two heterogeneous client bandwidth scenarios. In thescenario where different clients access the server via separate links, this paper proposes alocal search based progressive optimizing algorithm, which can greatly reduce thecomputational complexity of searching optimal unequal error protection solutions fordifferent client access bandwidths. In another scenario where many clients share a bottlenecklink, this paper proposes an embedded packetization framework for layered unequal errorprotection, and a performance evaluation function for it. Aimed at optimizing this function,this paper also proposes a local search algorithm and a fast heuristic algorithm. Experimentalresults show that, compared with the fixed-length packetization framework in the literature,the proposed embedded packetization framework can achieve much better client-weightedaverage performance.Thirdly, to address the quality smoothing problem encountered while streaming unequalerror protected scalable video over the Internet, this paper proposes an Additive Increase /Multiplicative Decrease quality control mechanism, and an online network adaptive bitallocation algorithm. To speed up this algorithm, a local search based progressive optimizingtechnique is adopted, and an approximate-approaching based dual-stage iteration technique isproposed. Experimental results show that, compared with the uniform-time bit allocationalgorithm, the proposed techniques can achieve near constant or graceful increasing quality ina segment by segment scheme, and the accelerated algorithm is very efficient and can be usedin real-time online applications.Finally, this paper implements an AVS streaming media demonstration platform whichcomplies with the ISMA (Internet streaming media alliance) specifications, and an AVS basedpeer-to-peer streaming media broadcast system: AVStreamer. Due to the heterogeneity ofnetworks, there may be random packet-drops while relaying media data from upstream peersto downstream peers. To address this problem, two priority based network adaptive andselective frame-drop algorithms are proposed. Analytical and experimental results show that,even in poor network conditions, the proposed algorithms are able to achieve much fluentrendering effect.As a conclusion, based on thorough analysis of some challenging problems in the area ofmedia streaming, this paper tries its best to improve the quality of service via application layertechniques, so as to provide robust and adaptive media streaming of the Internet. Theproposed thchniques have partially applied to practical systems and achieved favorableeffects.