Content-aware adaptation scheme for QoE optimized dash applications

The paper presents novel content-aware adaptation schemes for QoE optimized HTTP-based adaptive video streaming services. In recent years, content type has been investigated and proofed to have a deep influence on user's acceptable quality of experience (QoE). However, most of the existing HTTP-based adaptive video strategies only conduct bitrate switching according to network resources without a consideration of video content types. This may cause a non-optimal network resource allocation which may not achieve the optimized QoE for delivered video streaming services. In this paper, we proposed two content-aware adaptation schemes, named as Content-aware Probe and Adapt (C-PANDA) and Content-aware Dynamic Programming (C-DP), which are able to decide the video bitrate for the next video segment, based on not only the available bandwidth, the existing buffer capacity, but also the video content type. We implemented both content-aware adaptation algorithms in player prototypes based on libdash 3.0 library. Preliminary results show that content-aware adaptation schemes achieve better QoE when compared with conventional PANDA scheme. The proposed schemes are able to optimize QoE according to contents. They can achieve higher than acceptable QoE for most attractive contents such as those with high motion intensity to improve user experience for the most interesting scenes; select same video representations for segments belonging to a scene to avoid quality oscillation; and reserve buffer resource during low motion scenes to avoid stalling for the following high motion scenes. The work will be helpful in designing QoE-aware DASH schemes in the future.