Machine learning and IoT system for real-time cough detection and classification

Abstract

Our research investigates the application of machine learning and Internet of Things (IoT) technologies in healthcare, focusing on detecting and classifying coughing episodes. Leveraging deep learning architectures and a comprehensive IoT infrastructure, we developed an automated system capable of monitoring audio signals from a microphone array module to detect coughs and classify their types accurately. The study utilized the COUGHVID dataset for model training and evaluation, employing rigorous preprocessing techniques to ensure data integrity. Through comparative analysis, we identified MobileNet as the optimal model for cough detection, achieving promising results in accuracy, area under the ROC curve (AUC), and F1 score. Furthermore, our emphasis on privacy safeguards and remote medical examination facilitation underscores the practical implications of our research in enhancing healthcare delivery. Our study contributes to advancing technology-enabled healthcare solutions, offering valuable insights and solutions for improving patient care and outcomes.

Keywords

• machine learning
• mqtt
• cough detection
• healthcare
• deep learning
• spectrogram analysis