Early Detection of Respiratory Diseases Related to COVID-19 with Speech, Voice and Cough Analysis Software and Integration into Tele-Health Service

In the project, it is aimed to continuously monitor respiratory functions with speech, voice and cough analysis software and to detect respiratory tract disorders in the early period and integrate them into the telehealth system. Another aim of the project is to demonstrate the effectiveness of providing dietitian and psychological counseling services in an online system. Voice recordings taken from 25 patients with a positive diagnosis of COVID-19 and 25 patients with a negative diagnosis of COVID-19 were converted into 100 labeled data in total by separating speech and cough in preprocessing. Differences in the distribution of features (MFCC and VGGish) obtained from cough and speech recordings according to COVID-19 positive and negative classes were examined. The designed SVM classifier is trained separately with MFCC and VGGish features and the results are compared. Accordingly, the findings show that the VGGish attribute is successful in representing speech sounds and the MFCC attribute is successful in representing cough sounds. Face-to-face and online interviews of 30 patients with a dietitian and 20 patients with a psychological counselor were analyzed with questionnaires. In general, when both types of interviews are evaluated, there is no statistically significant difference between interview satisfaction and interview evaluation total scores of the participants who were interviewed face-to-face and online. The results of the analysis reveal the correlation of speech and cough sounds with COVID-19 and are of critical importance for the rapid detection of those with suspected disease. It has been seen that online interviews with psychologists and dieticians can be very useful, especially due to the conditions created by the pandemic period.