Modeling Best-Effort and FEC Streaming of Scalable Video in Lossy Network Channels

Video applications that transport delay-sensitive multimedia over best-effort networks usually require special mechanisms that can overcome packet loss without using retransmission. In response to this demand, forward-error correction (FEC) is often used in streaming applications to protect video and audio data in lossy network paths; however, studies in the literature report conflicting results on the benefits of FEC over best-effort streaming. To address this uncertainty, we start with a baseline case that examines the impact of packet loss on scalable (FGS-like) video in best-effort networks and derive a closed-form expression for the loss penalty imposed on embedded coding schemes under several simple loss models. Through this analysis, we find that the utility (i.e., usefulness to the user) of unprotected video converges to zero as streaming rates become high. We then study FEC-protected video streaming, re-derive the same utility metric, and show that for all values of loss rate inclusion of FEC overhead substantially improves the utility of video compared to the best-effort case. We finish the paper by constructing a dynamic controller on the amount of FEC that maximizes the utility of scalable video and show that the resulting system achieves a significantly better PSNR quality than alternative fixed-overhead methods