Automated cough detection system based on vision transformers

This study described the development of an automated cough detection system that utilized Vision Transformer architecture to improve cough detection accuracy. The trained model was evaluated using cough recordings from the RaDaR dataset, which comprised 24-hour acoustic cough recordings from patients from nine different diagnostic groups. All recordings were obtained using the Vitalograph VitaloJAK cough monitoring device, which captures cough sounds using a lapel microphone for ambient sounds and a contact microphone for thoracic sounds.