Page 148 - 《应用声学》2022年第4期

第 41 卷 第 4 期 Vol. 41, No. 4
2022 年 7 月 Journal of Applied Acoustics July, 2022

⋄ 研究报告 ⋄

利用卷积网络的高速列车主观声品质预测 ∗

贾尚帅 1 潘德阔 1 阮沛霖 2 郑 旭 2†

(1 中车唐山机车车辆有限公司技术研究中心 唐山 063035)
(2 浙江大学能源工程学院 杭州 310027)
摘要：随着高速列车在中国的高速发展，乘客对舒适性的要求不断提高。因此高速列车内声学舒适性是一个
需要研究和解决的问题。首先，该文基于声学人工头设备，获取了高速列车行驶在 350 km/h 速度下不同车厢、
不同区域的双耳噪声样本，并对其分别开展了主观声学评价和基于响度、尖锐度、粗糙度和抖动度等参数的客
观声品质分析。结果表明，350 km/h 速度下高速列车车内噪声能量集中在 3000 Hz 以内，风挡区域是舒适性
评价较差的位置，而响度是影响主观评价最大的因素。其次，利用卷积神经网络算法将主观评价结果与高速列
车噪声样本相关联，建立了车内噪声主观声品质预测模型，并与基于 BP 神经网络的预测模型进行了对比。结
果表明，基于卷积神经网络的主观声品质预测模型具有更高的预测精度，更适宜用于指导高速列车车内声学
舒适性的优化设计。
关键词：高速列车；车内噪声；声品质；卷积神经网络
中图法分类号: O429 文献标识码: A 文章编号: 1000-310X(2022)04-0646-08
DOI: 10.11684/j.issn.1000-310X.2022.04.017

Prediction of sound quality of high-speed train using convolutional network

JIA Shangshuai 1 PAN Dekuo 1 RUAN Peilin 2 ZHENG Xu 2

(1 Technical Research Center, CRRC Tangshan Co., Tangshan 063035, China)
(2 College of Energy Engineering, Zhejiang University, Hangzhou 310027, China)

Abstract: With the rapid development of high-speed train (HST) in China, passengers’ requirements for
comfort are increasing. Therefore, how to improve the interior noise and comfort of HST is a problem that needs
to be studied and resolved. Firstly, with the artificial head device, this paper obtains binaural noise samples
from different areas in different cabins of HST at a speed of 350 km/h. Then subjective evaluations and analysis
with loudness, sharpness, roughness, and fluctuation strength are carried out. The results show that the interior
noise energy of high-speed trains at 350 km/h is concentrated within 3000 Hz. The poorest evaluation result
appears at windshield area, and the loudness is the most important factor affecting the subjective evaluation.
Secondly, the convolutional neural network (CNN) is applied to build a sound quality prediction model between
the subjective evaluation results and the HST noise samples, and the model is compared with the prediction
model based on the BP neural network. The results show that the CNN prediction model has higher prediction
accuracy and can guide the optimization design of the HST.
Keywords: High speed train; Interior noise; Sound quality; Convolutional neural network

2021-06-16 收稿; 2021-11-29 定稿
国家自然科学基金项目 (51975515, 51705454)
∗
作者简介: 贾尚帅 (1982– ), 男, 河北唐山人, 博士, 中车资深技术专家, 研究方向: 高速列车振动噪声。
† 通信作者 E-mail: zhengxu@zju.edu.cn