Akademik Veri Yönetim Sistemi
8th International Conference on Smart City Applications, SCA 2023, Paris, Fransa, 4 - 06 Ekim 2023, cilt.906 LNNS, ss.457-466
COVID-19 is an upper respiratory disease that emerged in the last months of 2019 and affecting people worldwide. According to the World Health Organization data, it is determined that the outbreak spread rapidly. Early diagnosis is very important in preventing the spread of epidemics. Highly accurate identification and isolation of infected persons is essential to prevent the spread of outbreaks. Currently, reverse transcription polymerase chain reaction (RT-PCR) test is the most used method to detect COVID-19. However, the test is expensive, time-consuming, socially distant, difficult to distribute and has a high false negative rate. In this study, we developed a system based on Machine Learning (ML) and Deep Learning (DL) methods using voice signals (cough sounds) for early detection and diagnosis of COVID-19 disease. Since cough sounds are also a symptom of many upper respiratory tract diseases, a three-class system was developed to determine whether the voice recording from a person was COVID-19, upper respiratory tract disease, healthy or not. In this paper, indicative features were extracted Mel Frequency Cepstrum Coefficients (MFCC) from audio signals and features were classified with K-Nearest Neighbor Algorithm (KNN), Random Forest (RF), Decision Tree, Support Vector Machines (SVM), Logistic Regression classifiers. In addition, Convolutional Neural Networks (CNN) methods, one of the DL methods, was used and comparative performance results are given in the experimental section.