Real-time video streaming using prediction-based forward error correction

Real-time video streaming applications typically use an on-line forward error correction (FEC) technique to recover transmission losses with a low delay overhead. However, transmitting prioritized video data over variable-rate transmission channels complicates the FEC rate allocation process. Specifically, on-line FEC schemes result in an inefficient utilization of the available FEC bandwidth in the absence of prior information regarding the statistics of video traffic. In most streaming networks, the optimal FEC configuration is computed off-line in accordance with an analytical model. However, the present study proposes an on-line FEC scheme in which real-time FEC allocation is performed by extending the analytical FEC model with the frame size prediction technique. In the proposed approach, the optimal FEC configuration is computed in advance on a frame-by-frame basis over a series of predicted video frames, thereby yielding a significant reduction in the data buffering delay. The performance effects of frame-size prediction errors are mitigated by continuously revising the FEC configuration each time a new frame arrives. Moreover, a transmission rate control mechanism is proposed to ensure that each video frames satisfies its presentation deadline. The simulation results show that the proposed prediction-based FEC scheme minimizes the FEC processing delay while achieving virtually the same perceived video quality as that obtained using the off-line optimal FEC model.