Convolutional Neural Network-based Model for Lung Sounds Classification

Deep learning approaches are gaining popularity in the medical field for diagnostics and predictive analytics. This study aims to improve the classification of respiratory sounds recorded by electronic stethoscopes using a deep approach based on a convolutional neural network (CNN). The proposed scheme's performance is studied using the largest publicly available lung sound dataset (ICBHI 2017). We implement variant frequency representations, including CQT, STFT, and Mel-STFT, combined with Empirical mode decomposition (EMD). We propose the use of data normalization on top of data augmentation techniques. Furthermore, to increases the model performance, we used hyperparameters optimization and regulators. Experimental results show that our method outperforms competing methods on anomaly prediction tasks.