Is it worth to predict overflows during video streaming over wireless networks?

Abstract

No. In the paper, we consider compressed video streaming over the wireless channel. From time to time, dynamically changing channel characteristics cause queue overflows, packet drops and, finally, video quality degradation. To eliminate it, the well known I-Frame delay algorithm drops the least important packets when the queue overflows. However, this algorithm starts working only when the queue overflows. Inspired by many works on video bitrate and channel quality prediction, we study how I-Frame Delay would benefit if it knew about future queue overflows and filtered packets in advance. We show that even with the ideal prediction the improvement is negligible.