Advanced Contactless Coughing Monitoring Using Deep Learning and RF Technology

This study reports a contactless method for cough monitoring, employing a novel radio-frequency (RF) technology-based system. The primary objective is to develop a contactless system that can accurately recognize various cough types using deep learning algorithms integrated with RF technology. Method: The approach utilizes software-defined radios (SDRs) that emit and capture broad-spectrum RF signals at 5.6 GHz aimed at the subject's chest and lips. The communication between the SDRs uses orthogonal frequency division multiplexing (OFDM) signals, which are modulated by reflections from the subject's chest and lip movements during coughing. The received signal includes the channel frequency response (CFR), capturing the nuances of cough-induced movements. The collected raw CFR data is processed using deep learning and machine learning classifiers. A specialized dataset, developed from volunteer contributions, serves as the foundation for training these classifiers. The study introduces a blended Convolutional Neural Network-Gated Recurrent Unit (CNN-GRU) model, specifically designed to categorize five distinct coughing patterns: whooping cough, chronic cough, acute cough, subacute cough, and throat clearing. Results: The experimental setup involved 10 participants who simulated four types of abnormal coughing patterns and normal throat clearing. The data acquisition yielded 5,475 sets of 60 -second cough segments. The CNN-GRU model demonstrated superior performance in classifying cough types, achieving impressive metrics of 99.16 % accuracy, 99.39 % precision, 99.25 % recall, and a 99.52 % F1-score. Conclusion: The system's ability to accurately distinguish between different coughing patterns enhances its potential to aid the diagnosis of various respiratory conditions. This RF-based non-contact cough monitoring system ensures uninterrupted, ongoing monitoring of cough trends, marking a substantial advancement in the field of respiratory health monitoring.