Citation

Sreejith, Reshma and Ramasamy, R. Kanesaraj and Mohd Isa, Wan Noorshahida and Abdullah, Junaidi and Md. Jamal, Shamsuriani and Amri Hamzah, Faizal and Thanjappan, Sivasutha (2026) RespireNet: Enhancing Lung Sound Classification using CNN-TCN Hybrid Approach. In: AIAT 2025: 2025 5th International Conference on Artificial Intelligence and Application Technologies, Kyoto, Japan, 4-6 December 2025.

Text 4.pdf - Published Version Restricted to Repository staff only Download (610kB)

Abstract

Respiratory sound analysis plays a vital role in the early detection and diagnosis of pulmonary diseases such as asthma, COPD, and pneumonia. Traditional auscultation, while widely practised, is often subjective and limited in its ability to capture subtle abnormalities. To address these challenges, this study introduces RespireNet (CNN–TCN), a hybrid deep learning model that combines Convolutional Neural Networks (CNNs) for spectral feature extraction with Temporal Convolutional Networks (TCNs) for modeling long-range temporal dependencies in respiratory cycles. The preprocessing stage employs Discrete Wavelet Transform (DWT) denoising to reduce background noise, followed by MFCC-based feature extraction for robust representation of lung sounds. Evaluated on the ICBHI 2017 dataset, RespireNet achieved an accuracy of 95.49%, a precision of 94.80%, recall of 94.20%, and an F1-score of 94.49%, substantially outperforming baseline models such as CNN, CNN+SVM, CNN+RNN, CNN+LSTM, and CNN+BiGRU. Confusion matrix analysis further demonstrated that RespireNet consistently minimised misclassifications across the four sound categories, confirming its robustness and reliability. These findings establish RespireNet as a promising framework for automated lung sound classification, with strong potential for integration into intelligent stethoscopes and telemedicine

Item Type:	Conference or Workshop Item (Paper)
Uncontrolled Keywords:	Lung sound classification, Respiratory sound analysis, Deep learn ing, Convolutional Neural Network (CNN), Temporal Convo lutional Network (TCN), Discrete Wavelet Transform (DWT), RespireNet
Subjects:	Q Science > Q Science (General)
Divisions:	Faculty of Computing and Informatics (FCI)
Depositing User:	Ms Suzilawati Abu Samah
Date Deposited:	02 Apr 2026 04:38
Last Modified:	06 Apr 2026 08:31
URII:	http://shdl.mmu.edu.my/id/eprint/15655

Downloads

Edit (login required)