Rate control optimization for live streaming using fixed-time playback buffer

Abstract

Rate control plays an important role in video encoding. The complexity of a video for each frame changes and it cannot be predicted accurately. However, there are always strict rules for live encoded video stream and it has different limits for the bitrate and range of fluctuation under various application scenarios. So, rate control is a challenging research. Traditional rate control models such as hypothetical reference decoder and video buffer verifier have rigid requirements for video players and are not adaptable to mainstream players and live streaming services. Here, we propose an optimization method of rate control for live streaming. In this paper, firstly, we propose a playback strategy that uses a fixed-time buffer, as opposed to a data volume buffer, in conjunction with a corresponding buffer model. Secondly, we propose a judgment method that can determine whether the video hangs or pauses unexpectedly using the playback strategy. We also derive an easy-to-implement and low-complexity algorithm and realize optimization for live streaming applications. Thirdly, the forecast-based rate control algorithm for live video streaming is given, which is consistent with the judgment method. In addition, we develop a quality optimization of the rate control algorithm based on viewer experiences. Lastly, our empirical analysis and experiments verify the effectiveness of the proposed judgment method and rate control algorithms. The proposed methods and algorithms are also compatible and implementable in various players, including Adobe Flash, iOS video player, and VLC.