Improving Optimization-Based Rate Adaptation in Dash System

Bo Wang,Xiaohui Luo,Ping Hu,Fengyuan Ren

DOI: https://doi.org/10.1109/icccn.2017.8038402

2017-01-01

Abstract:More and more commercial video players use bitrate adaptation to adjust video quality according to varying network conditions. Optimization-based approaches are widely used for bitrate adaptation in Dynamic Adaptive Streaming over HTTP (DASH). Essentially, the optimization problem is solved based on the prediction of buffer dynamics. However, stochastic chunk size deviates observably the buffer occupancy from the expected value, making the evolution hard to predict. In order to get rid of this effect and improve the prediction accuracy for buffer occupancy, we propose an algorithm based on markov decision process with incorporating chunk size information so that only the network capacity variation need to be considered in the decision-making process. Experiment results show that our solution can effectively eliminate performance oscillation induced by variable chunk size and achieve a good QoE.