A Content-Aware Deep Q-Learning Approach for Adaptive Video Streaming
Hala Amer, Mohamed S. Hassan, Mahmoud H. Ismail · 2024
Adaptive streaming over HTTP aims to maximize user Quality-of-Experience through video quality adaptation. Conventional adaptation schemes measure the video quality for variable bitrate (VBR) video in terms of the average bitrate. However, video bitrate is not an accurate measure of perceptual quality. Alternative quality measures, such as the Video Multi-method Assessment Fusion (VMAF), can be used to better represent the quality perceived by the viewer. Studying the VMAF of video chunks across the same bitrate level shows that user QoE depends not only on the overall video quality, but also on the content complexity. This work proposes a deep Q-learning (DQL) adaptation algorithm that accounts for content complexity by prioritizing complex video chunks during bitrate selection. Simulation results show that the average video quality is improved by using complexity-aware streaming over the baseline algorithms,