A MEDICAL MOBILE APPLICATION FOR COVID-19 DIAGNOSIS USING COUGH SOUND

Abueisha, Ibrahim Reda Ibrahim; Abid, Samah Bayoumi; Kishk, Salma Mohamed; Salem, Zeinab Galal Yahya

doi:10.21608/iugrc.2022.302673

A MEDICAL MOBILE APPLICATION FOR COVID-19 DIAGNOSIS USING COUGH SOUND

Document Type : Original Article

Authors

Tanta University, Egypt.

10.21608/iugrc.2022.302673

Abstract

Since the COVID-19 pandemic started, researchers have indicated potential techniques to develop COVID-19 screening tools. One practical and affordable solution is to use cough recordings for COVID-19 detection. Based on the combination of Deep learning and signal processing approaches, we present a mobile application for COVID-19 detection using cough recordings. First, AI model is developed and trained using the COUGHVID cough dataset. Our model uses convolution neural networks and an image classifier, to identify COVID-19 infection from an audio file. It takes a Mel Scale spectrogram as an input, which is an image representation of the audio stream and classifies it into COVID infected or healthy. The testing accuracy for our classification model was 98.7%. Then, we develop a mobile application to receive the cough recoding from the user and display the result. Our application is connected to a server which receives the audio file from the application as an HTTP request, runs the stored Python code on it, and then returns a result as an HTTP response

Keywords