高级卷积神经网络在多分类呼吸音分析中的应用

Application of Advanced Convolutional Neural Networks in Multiclass Respiratory Sound Analysis

吴俊¹

扫码查看

作者信息

1. 蛮牛健康管理服务有限公司,上海 200040
折叠

摘要

文章旨在构建和评估基于深度学习的肺呼吸音分类器,用于区分支气管炎、慢性阻塞性肺炎、肺炎、上呼吸道感染和健康状态.采用ICBHI 2017公开数据集,通过数据增强技术和频谱特征提取,构建卷积神经网络(CNN)、深度神经网络(DNN)和残差网络(ResNet)3种模型.实验结果表明,这些模型在呼吸音分类任务中表现出色,其中,ResNet模型在所有模型中表现最佳.

Abstract

This paper aims to construct and evaluate deep learning-based lung sound classifiers to differentiate bronchitis,chron-ic obstructive pulmonary disease(COPD)pneumonia,upper respiratory tract infection(URTI),and healthy states.Using the publicly available ICBHI 2017 dataset,this paper employs data augmentation techniques and spectral feature extraction to build three models,convolutional neural network(CNN),deep neural network(DNN)and residual networks(ResNet).The exper-imental results demonstrate that these models perform excellently in the task of lung sound classification,with the ResNet model showing the best performance across all models.

关键词

深度学习/卷积神经网络/深度神经网络/残差网络

Key words

deep learning/convolutional neural network/deep neural network/residual network

引用本文复制引用

出版年

2024

微型电脑应用

上海市微型电脑应用学会

微型电脑应用

CSTPCD

影响因子：0.359

ISSN：1007-757X

段落导航