摘要： 针对IP网络质量评价问题，利用BP神经网络构建一种主客观特征融合的评价模型。主客观2个评价网络分别进行自适应学习，在学习输出稳定后，2个BP神经网络通过相互共振学习使主客观评价数据融合，由此获得一致的评价结果。结合用户感知的主观评价和多种客观性能评价指标对该模型进行仿真，实验结果表明，模型具有较强的抗噪性能，可以较好地满足IP网络质量的综合评价要求。

关键词: IP网络质量, 评价模型, BP神经网络, 主客观评价融合, 共振学习

Abstract: Aiming at quality evaluation of IP network problem, this paper proposes an evaluation model of subjective and objective fusion by using Back Propagate(BP) neural network. The two neural networks of subjective and objective characteristic are self-adapting learned separately until the output is stable. The two BP neural networks are resonated mutually, and the subjective evalution and objective evalution are fused each other by the resonance learning of two networks. An accordant evaluation is obtained after the resonance study. Simulations of users subjective evaluation and many objective indexes of IP network are practiced, which indicate that the fusion model in this paper has stronger performance of noise resistance. The fusion model can better satisfy the composite requirements of IP network quality evaluation.

Key words: IP network quality, evaluation model, Back Propagate(BP) neural network, fusion of subjective and objective evaluation, resonance study

中图分类号: 

• TP393