摘要
为有效感知网络视频末端受众体验质量,提出一种利用网络层QoS指标数据在线实时评估网络视频QoE的方法。基于BP神经网络,分别使用变学习率梯度下降和弹性BP两种算法,建立关系映射模型。在搭建仿真实验环境,模拟视频服务端到端传输过程和网络损伤,获取QoS与QoE指标数据的基础上,进行模型误差分析,实验结果表明,映射模型所得到的预测数据,与目标数据分布特征规律基本吻合,精确度满足预期要求。
For the perception of end audience＇s subjective feeling quality, one kind of BP-based estimate on network video quality of experience （QoE） was presented. Based on the BP neural networks, the variable learning rate gradient descent algorithm and stretch BP algorithm were used to construt the mapping model. Through building experiment environment and simulating the transmission of network video and network damage, the QoS and QoE were fetched, and the model error analysis was made. Experimental result shows that the predicted data obtained from the model, its distribution characteristics are basically consistent with the target data, accuracy meets the expected requirements.
出处
《计算机工程与设计》
北大核心
2017年第1期1-6,共6页
Computer Engineering and Design
基金
国家863高技术研究发展计划基金项目(2015AA020101)
