Identification of COVID-19 from Cough Sounds Using Non-Linear Analysis and Machine Learning

Automatic diagnosis of COVID-19 has an active role in reducing the spread disease by minimizing interaction with people. Machine learning models using various signals and images form basis automatic diagnosis. This study presents machine based for detecting infection ‘Virufy’ dataset containing cough sound labeled as Non-COVID-19. Since number COVID positive coughs set is less than those negative, firstly, data balancing was performed ADASYN oversampling technique study. Then, features were extracted non-linear analysis sounds Multifractal Detrended Fluctuation Analysis (MDFA), Lempel–Ziv Complexity (LZC) entropy measures. Later, most effective selected ReliefF method. Finally, five algorithms, namely Support Vector Radial Basis Function (SVM-RBF), Random Forest (RF), Adaboost, Artificial Neural Network (ANN), k Nearest Neighbor (kNN) used to identify or Non-COVID19. As a result study, patients Non-COVID19 subjects identified 95.8% classification accuracy thanks RBF kernel function SVM features. With this classifier, 93.1% sensitivity, 98.6% specificity, precision, 0.92 kappa statistical values 93.2% area under ROC curve obtained.