基于MobileNetV3卷积神经网络的供水管道漏损音频分类  

Audio Classification of Water Supply Pipeline Leakage Based on MobileNetV3 Convolutional Neural Network

作　　者：陈双叶 徐雷桁 黄成意 张智武 张林 韩默 CHEN Shuangye;XU Leiheng;HUANG Chengyi;ZHANG Zhiwu;ZHANG Lin;HAN Mo(Faculty of Information Technology,Beijing University of Technology,Beijing 100124,China;Beijing Water Supply Group Yutong Municipal Engineering Co.,Ltd.,Beijing 100142,China)

机构地区：[1]北京工业大学信息学部,北京100124 [2]北京市自来水集团禹通市政工程有限公司,北京100142

出　　处：《北京工业大学学报》2024年第7期797-804,共8页Journal of Beijing University of Technology

基　　金：国家重点研发计划资助项目(2017YFB0306404)。

摘　　要：为了对城市供水管网漏损音进行准确识别,提出一种基于MobileNetV3的供水管道漏损音频分类识别方法。首先将ROPP数据集中的音频文件进行离线数据增强,将漏损信号转变为对数梅尔谱图并采用谱减法实现数据降噪;然后使用注意力机制模块与MobileNetV3网络训练识别并提取图像特征;最后使用Softmax函数对漏损音频进行分类。实验结果表明,该方法可以使漏水类别的分类精确度达到99.40%,召回率达到99.20%。To accurately identify the leakage sound of the urban water supply network,this paper proposes an audio classification and recognition method for water supply pipeline leakage based on MobileNetV3.First,the audio files in the ROPP dataset were enhanced offline,the leakage signal was converted into a logarithmic Mel spectrogram and the spectral subtraction was used to achieve data noise reduction.The Attention Mechanism module and MobileNetV3 neural network training were used to extract image features.Finally,the Softmax function was used to classify the leaked audio.The experimental results show that this method can make the classification precision of water leakage category reach 99.40%,and the recall rate reaches 99.20%.

关 键 词：声音事件分类 水管泄漏检测 MobileNetV3 数据增强 谱减法 压缩奖惩网络模块 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] TP273[自动化与计算机技术—控制科学与工程]