Cough Sound Based COVID-19 Detection System Using Machine Learning Algorithms

الملخص

Due to the nature of the COVID-19 pandemic, the need for early detection is
essential for a rapid recovery and limiting the spread of the virus. Ordinary
and traditional methods of diagnosing COVID-19 depend on contact which is
susceptible to transmission of the virus, as any misuse of traditional methods
can lead to the spreading of the epidemic, increasing its severity and may lead
to death. However, since the better methods and techniques are always
required for diagnosis, this study is aimed at presenting a contactless approach
to distinguish COVID-19 infection from other similar symptoms infections
based on the cough sound. COVID-19 detection system has been implemented
by using machine learning techniques; the Mel Frequency Cepstral
Coefficients (MFCC) algorithm for extracting features from audio signals and
Multilayer Perceptron Neural Network (MLP) for classification. The system
has been implemented by using a sample of data downloaded from online
provided COUGHVID dataset. The model has considerably shown a high
performance as it achieved 96%, 92%, and 100% for average accuracy,
sensitivity and specificity respectively