QoS Control For Video Streaming Over Internet And Wireless Networks

The emerging national audio/video coding standard AVS 1.0, the availability of high bandwidth Internet access and the forthcoming deployment of 3G networks together lay the foundation of video streaming over Internet and wireless networks, but the existing multimedia streaming systems must be improved in order to provide satisfactory performance. This thesis proposed a novel rate control scheme for video streaming over wireless, and put forward a method to encapsulate AVS video stream into a MP4 file, which makes possible the streaming of AVS video over IP networks.In this thesis we first reviewed the main components of a general multimedia streaming system, and pointed out the necessity and importance of MP4 encapsulation of AVS. The application layer QoS control module was reviewed in detail.Second, we briefly introduced the MP4 file format and its organization, and explained the relation between AVS Video Track and Hint Track. The method to encapsulate AVS video stream into a MP4 file was presented in detail, and the process of creating AVS Video Track and Hint Track was particularly emphasized.Then we focused on the research on rate control scheme for video streaming over wireless. TFRC was the widely accepted rate control scheme in wired IP networks, but it suffered performance degradation in wireless networks. We reviewed the current solutions to this problem, analyzed their limitation, and proposed our novel scheme LERD. LERD divides each TFRC connection into multiple discounting levels. Each level is characterized by a unique discounting factor to revise the untruly high loss event rate. The proper discounting level of different network situation was determined by IIAD algorithm based on the feedback information provided by the network. Then we gave the experimental results of TFRC-LERD and TFRC and explained the reason. The result indicated that TFRC-LERD could detect the bandwidth of wireless link more precisely, and it was TCP friendly.Finally we applied LERD to MULTFRC and achieved significant performance improvement. The experimental results demonstrated that MULTFRC-LERD outperformed MULTFRC in several aspects. Specifically speaking, it reduced the resource consumption, extended the area of applicability and mitigated the quantization effect at the low error rate environment.