基于RBF神经网络的车内声品质预测及分析

The Vehicle Interior Sound Quality Prediction and Analysis Based on RBF Neural Network

针对特种车车内声品质,提出了一种关于声品质客观评价参数对主观评价结果预测和影响权重的径向基函数(RBF)神经网络方法。首先对三辆不同类型特种车进行实车道路试验,用等级评分法进行主观评价试验,计算声品质的客观评价参数;然后搭建关于车内声品预测的RBF神经网络模型,以客观评价参数作为RBF神经网络输入,主观评价结果作为输出,通过仿真与声品质的主观评价值对比获得较好的一致性;最后利用各网络层间的连接权值,计算车内声品质客观评价参数对主观烦躁度的影响权重。探究表明:车内声品质主要受到总响度、响度和语言干扰级(SIL-4)3个客观参数的影响。

Considering the sound quality of the special vehicle,a radial basis function(RBF)neural network method is proposed for calculating the weight and predicting subjective evaluation results concerning the influence of the objective evaluation parameters.First,three different types of special vehicles were tested on the road,and a subjective evaluation test was carried out to calculate the objective evaluation parameters of the sound quality by using the grade method.Then,the RBF neural network model was established for sound quality prediction inside the vehicle.The objective evaluation parameters are regarded as the input of RBF neural network model,and the subjective evaluation results as output,and the good consistency is obtained by comparing the simulation result with the subjective evaluation value of sound quality.Finally,the connection weight between the network layers was applied to calculate the influence weight of the objective evaluation parameters on the subjective irritability.The results show that the sound quality of the vehicle is mainly affected by the objective parameters of the overall loudness,loudness and speech interference level(SIL-4).