Lung sounds classification using convolutional neural networks
作者：
Highlights：
• We have proposed an approach for lung sounds classification using CNNs and compared it with two features-based approaches.
• In the first handcrafted features-based approach, we used MFCC’s statistics extracted from the signals.
• In the second one, we used local binary patterns extracted from spectrograms.
• We also opted for ‘Handcrafted Features + CNN’ and ‘CNN Features + Classifier’ configurations.
• The final results show that CNNs outperformed the handcrafted features-based approaches.
摘要
•We have proposed an approach for lung sounds classification using CNNs and compared it with two features-based approaches.•In the first handcrafted features-based approach, we used MFCC’s statistics extracted from the signals.•In the second one, we used local binary patterns extracted from spectrograms.•We also opted for ‘Handcrafted Features + CNN’ and ‘CNN Features + Classifier’ configurations.•The final results show that CNNs outperformed the handcrafted features-based approaches.
论文关键词：Convolutional neural network,Lung sounds classification,Handcrafted features extraction,Deep learning,Models ensembling,Support vector machines
论文评审过程：Received 5 May 2017, Revised 18 April 2018, Accepted 23 April 2018, Available online 1 May 2018, Version of Record 7 June 2018.
论文官网地址：https://doi.org/10.1016/j.artmed.2018.04.008
