人群差异对多维度声品质评价的影响  

Influence of population differences on multidimensional sound quality evaluation

作　　者：吴启斌 王国盛 赵向阳 刘宇珂 WU Qibin;WANG Guosheng;ZHAO Xiangyang;LIU Yuke(Department of Automobile Engineering,Henan Institute of Technology,Xinxiang 453003,Henan,China)

机构地区：[1]河南工学院车辆与交通工程学院,河南新乡453003

出　　处：《声学技术》2025年第2期252-260,共9页Technical Acoustics

基　　金：河南省高等学校重点科研项目(22A460008);河南省科技攻关项目(252102240107)。

摘　　要：为建立基于多维度的某国产运动型多用途汽车车内声品质评价模型,首先,采集车内驾驶员处的稳态和非稳态噪声样本信号,组织噪声评价的评审团,并通过调研确定评审员用户画像,将声品质分为舒适感和运动感两个维度,采用等级评分法进行不同维度下的声品质主观评价;然后,基于Matlab计算出样本信号的声品质客观参量,并与主观评价得分进行相关性分析,确定与舒适感和运动感相关性强的参数;最后,将样本信号分为训练样本和验证样本,并将训练样本的客观参量和主观评价得分作为声品质客观评价模型的输入和输出参数,基于Matlab Simulink建立遗传算法-前馈型(genetic algorithm-back propagation,GA-BP)神经网络评价数学模型。经过验证样本的检验,GA-BP模型对于两个维度的声品质评价误差最小,预测误差平均值分别为3.63%和3.04%。In order to establish a multidimensional evaluation model for the interior sound quality of a domestic sport utility vehicle(SUV),steady-state and non-stationary noise sample signals are first collected from the driver's position in the car,and a noise evaluation review panel is organized.Through research,the user profile of the reviewer is determined,and sound quality is divided into two dimensions:comfort and sportiness.Subjective evaluations of sound quality under different dimensions are conducted using a grading method.Then,based on Matlab,the objective parameters of the sound quality of the sample signals are calculated and correlated with the subjective evaluation scores to determine the parameters strongly correlated with comfort and sportiness.Finally,the sample signals are divided into training samples and validation samples,and the objective parameters and subjective evaluation scores of the training samples are used as input and output parameters for the objective evaluation model of sound quality.A genetic algorithm-back propagation(GA-BP)neural network mathematical model is established using Matlab Simulink.Through verification of the samples,the GA-BP model has the smallest error in evaluating sound quality in both dimensions,with average prediction errors of 3.63%and 3.04%,respectively.

关 键 词：声品质 多维度 用户画像 遗传算法-前馈型神经网络评价模型 

分 类 号：U461.4[机械工程—车辆工程]