Covid-19 Detection from Cough, Breath, And Speech Sounds with Short-Time Fourier Transform and a CNN Model

Ekiz A., KAPLAN K.

2023 Innovations in Intelligent Systems and Applications Conference, ASYU 2023, Sivas, Türkiye, 11 - 13 Ekim 2023 identifier

  • Yayın Türü: Bildiri / Tam Metin Bildiri
  • Doi Numarası: 10.1109/asyu58738.2023.10296675
  • Basıldığı Şehir: Sivas
  • Basıldığı Ülke: Türkiye
  • Anahtar Kelimeler: CNN, COVID-19, deep learning, Short-time Fourier transform, sound analysis
  • Kocaeli Üniversitesi Adresli: Evet

Özet

To eliminate the negative effects of existing methods such as social distance rule violation, slow test time and to create a pre-diagnosis method, deep learning and sound analysis work has been carried out for the Covid-19 disease, which has turned into a pandemic. For this purpose, experiments were performed on the crowdsourced Coswara dataset for Covid-19 Detection from Cough, Breath and Speech Sounds with Short-Time Fourier Transform and a CNN Model. On Coswara dataset, Covid-19 tested samples were selected and the CNN model was trained on different type of sounds. The best result was achieved with cough-heavy sound type as 0.980 precision, 0.998 AUC, 0.990 F1-score on test set.