COVID-19 Detection in Cough Audio Dataset Using Deep Learning Model

Using Deep Learning methods might be a proper answer to the need of the world for a fast, automatic solution for COVID-19 early-stage diagnosis. This article tries to take advantage of Convolutional Neural Network (CNN) systems for this purpose. Our proposed model is based on a CNN network and is trained based on the COUGHVID dataset. By implementing feature extraction using MFCC and using data augmentation methods, we tried to develop a fully functional model. The results show there were improvements compared to other state-of-the-art projects. Based on the metrics used in this work, we achieved an area under the curve of the receiver operating characteristics (AUC-ROC) of 0.94 on the task of COVID-19 classification. © 2022 IEEE.