QUVE: QoE Maximizing Framework for Video-Streaming

As video-streaming services for mobile terminals are becoming more popular, the volume of mobile traffic is growing rapidly. To deliver traffic-heavy network services without degrading users’ overall experience, one must examine service characteristics and end-to-end network conditions. In this paper, we propose “QUVE,” a framework for maximizing the user's quality of experience (QoE) of video streaming services. The QUVE framework consists of two key components: a QoE estimation model and QoE parameter estimation method. The QoE-estimation model is based on the rebuffering count and time, and content-encoding conditions. The QoE parameter-estimation method estimates forthcoming network quality and the corresponding rebuffering count and time that the user will experience. The effectiveness of this framework was demonstrated through a large-scale field trial for Niconico video service, one of the most popular video-streaming services in Japan. We gathered more than 1.4 billion pieces of feedback data from the in-service trial and found that our framework enhances user QoE by selecting the best encoding conditions suited for user network conditions.