计算机工程

• •    

C-ABR:一种基于视频分类的码率自适应算法

  

  • 发布日期:2020-12-18

C-ABR:An adaptive bitrate algorithm based on video type classification

  • Published:2020-12-18

摘要: 流媒体的码率自适应算法依据网络状态动态调节视频块的码率,以提升用户体验质量。然而目前的码率自适应算法设计忽略了视频类型的差异对用户体验质量的影响,导致性能下降。本文不同视频类型设计了相应的用户体验质量效用函数,使用强化学习算法训练模型A3C,实现视频类型相适应的码率自适应算法C-ABR。实验结果说明,相对于目前典型的码率自使用算法Pensieve和MPC, C-ABR方法将用户体验质量分别提升了22.7%和50.4%。

Abstract: The bit rate adaptive algorithm of streaming media dynamically adjusts the bit rate of the video block according to the network status to improve the quality of user experience. However, the current rate adaptive algorithm design ignores the impact of the difference of video types on the quality of user experience, resulting in performance degradation. In this paper, the corresponding user experience quality utility functions are designed for different video types, and the model A3C is trained using the reinforcement learning algorithm to realize the bit rate adaptive algorithm C-ABR adapted to the video type. The experimental results show that, compared with the current typical self-use algorithms Pensieve and MPC, the C-ABR method improves the user experience quality by 22.7% and 50.4%, respectively.