M.G. Manisha Milani, Murugaiya Ramashini, Pg Emeroylariffion Abas


Coronavirus or COVID-19 has been declared a global pandemic, spreading to most countries, infecting millions of people and causing hundreds of thousands of deaths worldwide. The virus directly affects the respiratory system and causes many symptoms, including fever, shortness of breath and cough, and in more severe cases, it can cause pneumonia. Early detection of COVID-19 infection has become essential to curb the spread of the virus. With cough appearing as one of the common symptoms, cough sound may be used as tools to assist in diagnosing COVID-19 individuals. This paper extracts Mel Frequency Cepstral Coefficient (MFCC) and Gammatone Cepstral Coefficients (GTCC) from individual’s cough sounds and use them to classify the sounds via an ensemble bagged tree supervised classification model to differentiate COVID-19 and healthy individuals. An online database containing cough sounds of individuals with and without COVID-19 has been used to test the model. It has been shown that classification accuracies are higher for model employing GTCC features, with 87.5% and 75% accuracy for COVID-19 and healthy individuals, respectively. Increasing the training capacity of the proposed classification model may provide even higher accuracy. The result is significant, as the system can be developed into a real time coronavirus detection system to assist medical practitioners in diagnosing potential COVID-19 patients and limit the spread of the contagious virus.

Full Text:



  • There are currently no refbacks.