Network adaptive packet scheduling for streaming video over error-prone networks

In packet networks including the Internet and commercial 3G wireless bearers, network conditions that a streaming application experiences are not known a priori and exhibit time-varying characteristics. For such environments, network-adaptive techniques are important to efficiently stream video. We propose an adaptive packet scheduling algorithm for streaming video over error-prone networks. A MAP (maximum a posteriori) framework is incorporated for adapting to time-varying network loss rate. By MAP-based dynamic network estimation, the proposed scheduler adaptively performs video packet transmissions over time-varying networks. Also, packet transmission schedules are optimally determined in a distortion-optimized way. For low computational complexity, the proposed packet scheduling algorithm is based on a new and simple distortion measure which is referred to the EFS (expected frame-rate reduction). Simulation results for standard video sequence show that the proposed scheduling outperforms conventional ARQ-based schemes from a view point of frame loss rate and reconstructed video quality. In particular, the proposed scheduling algorithm shows significant improvement over highly lossy channel.