One way to reduce, or avoid, the loss of intrastream synchronization due to the delay variability introduced by best-effort networks, is by employing application layer buffering and scheduling at a Packet Video Receiver (PVR), resulting in a higher end-to-end delay. In this paper an analytical model is presented that captures the essential tradeoff between stream continuity and stream latency. Unlike past related work, stream continuity is not expressed as the accumulated amount of synchronization loss, but as a combination of the accumulated amount, and the variation of the duration of synchronization loss occurrences. This approach allows for a fine grained optimization of stream continuity which has the potential of providing an improved perceptual quality. It is shown that the minimization of the accumulated amount of synchronization loss, and the minimization of the variance of the duration of synchronization loss occurrences, are two competing objectives; the minimization of variance is desirable because it leads to the concealment of discontinuities. The aforementioned presentation quality metrics are considered by the optimal playout policy, which is derived by means of Markov decision theory and linear programming.
展开▼
