基于U-Net结构的噪声源定位分类方法研究  

Research on Noise Source Localization and Classification Method Based on U-Net Structure

作　　者：黄力 HUANG Li(Fuzhou Vocational and Technical College,Fuzhou 350108,China)

机构地区：[1]福州职业技术学院,福建福州350108

出　　处：《电声技术》2024年第12期51-53,共3页Audio Engineering

摘　　要：针对噪声污染控制中的噪声源精确定位与分类问题,提出一种基于U-Net结构的方法。该方法通过结合卷积神经网络的编码器-解码器架构与跳跃连接,能够在复杂声场中有效提取多尺度声学特征,实现高精度的噪声源定位和分类。实验结果表明,该方法在不同信噪比(Signal Noise Ratio,SNR)环境下都表现优异,为噪声污染治理提供了有力的技术支撑。Aiming at the problem of accurate location and classification of noise sources in noise pollution control,a method based on U-Net structure is proposed.This method can effectively extract multi-scale acoustic features in complex sound field by combining the encoder-decoder architecture of convolutional neural network with jump connection,and realize high-precision noise source location and classification.The experimental results show that this method performs well in different Signal Noise Ratio(SNR) environments,which provides strong technical support for noise pollution control.

关 键 词：噪声源定位 U-Net 深度学习 声谱图 

分 类 号：TB535[理学—物理]