基于BP神经网络QoS到QoE映射模型

QoS-QoE mapping model based on back propagation neural network

随着互联网的发展,基于语音和视频的网络应用服务层出不穷,人们开始不仅仅满足于知道网络QoS参数,更加关注网络服务的好坏,即QoE评价指标。目前常用的视频QoE评价的方法是基于图像评价的算法如PSNR算法、VQM评价算法等。这些算法需要原始视频图像进行对比,较为复杂,实时性差。本文研究了一种基于BP神经网络的QoS到流媒体QoE映射模型,使用抖动和丢包两个网络QoS参数作为输入层神经元,基于VQM算法的QoE评价值作为输出层神经元。使用若干组QoS和QoE数据训练该BP神经网络,分别使用单隐含层BP神经网络和多隐含层BP神经网络进行效果对比,之后使用该神经网络和QoS参数对QoE评价值进行预测,并与QoE实际值比较验证。得到基于BP神经网络的QoS到QoE的映射模型。该模型较为简单,拟合度高,RMS误差较小。

With the development of the Internet.People care more about the QoE（Quality of Experience）score of these services instead of the network parameter.While the QoE assessment of IPTV service based on image analysis like PNSR and VQM need full reference video to calculate the MSE with poor real time performance and also complicated.This paper had a research on the QoS to QoE mapping model based on the BPANNs,including the training and predict the QoE assessment using QoS parameter,then verification this model.Turn out that this QoS-QoE mapping model is well performed with high accuracy and fitting index.