Internet多媒体分层组播JSCC差错控制率失真优化

针对Internet多媒体群组通信中同时存在的带宽异构性和包丢失率异构性，文中将分层组播和接收者驱动的思想扩展到FEC差错控制中，提出一种分层FEC组播差错控制方法LM-FEC．LM-FEC通过不同的组播组发送信源编码层和各信源层的FEC校验数据，为接收者根据信道带宽和数据包丢失率实施差错控制提供更加灵活的选择．文中用FH-MDP模型描述接收者行为，通过JSCC率失真优化确定编码层内和编码层间的速率分配，JSCC率失真优化采用变量替换和动态规划算法求解．实验表明，该文提出的差错控制方法能够有效改善重建多媒体信号的回放质量．

RD Optimization in JSCC-Based Error Control on Internet Multimedia Layered Multicast

Because of the heterogeneities of Internet, error control in multimedia multicast communications over the Internet is much more complicated than that in end-to-end context. Normally the bandwidth heterogeneity and packet loss rate heterogeneity in multimedia group communications based on IP multicast technology are studied separately. This paper aims at the concurrence of Internet heterogeneities of bandwidth and packet loss rate and presents a new error control scheme LM-FEC for Internet multimedia group communications, by combining the idea of layered multicast and receiver-driven into FEC-based error control technology. In LM-FEC, layered multimedia source packets and FEC verifying packets of each source layer are multicast by the sender in different groups, therefore the receiver can implement error control on a flexible receiver-driven way. According to the available bandwidth and packet loss rate of the channel, which may change dynamically, and the real-time requirement from multimedia information, the receiver enters into or quits from some of multicast groups to obtain the reconstruction multimedia information with quality as best as possible. In this way, LM-FEC deals with unitively the bandwidth heterogeneity and packet loss rate heterogeneity in Internet multimedia group communications. With the receiver-driven way, LM-FEC doesn't need any feedback information. Thus means no feedback channel is required and the feedback implosion problem to which feedback based error control scheme is prone can be avoided totally. Receiver's action determines the inter layers and intra layers rate allocation of the available bandwidth. In this paper, an FH-MDP model is used to describe receiver's action. Receiver decides its policy with JSCC rate-distortion optimization. The optimization is solved with variable alternation scheme and dynamic programming algorithm. Simulative experiments are implemented to evaluate the performance of LM-FEC compared with PET and RLM in Internet multimedia multicast communications. The results show considerable gains on reconstructed video quality can be obtained with LM-FEC.