基于实时视频QoE的优化算法研究

How to improve the QoE (Quality of Experience) of low latency video transmission has been a challenging problem. Due to their defined rules, traditional transmission algorithms are difficult to cope with dynamic network environment changes in real-time video transmission. Emerging reinforcement learning-based transmission algorithms can improve bandwidth utilization, but the performance is not stable enough. In this paper, we deeply integrate the two types of algorithms through the self-attention mechanism Non-Local, which fully extracts the features of both types of algorithms and predicts more appropriate behaviors to improve the QoE. Compared with the method of setting a fixed bitrate in the transmission process, the research method in this paper has improved in various metrics and has further research significance.