Enhancing Respiratory Disease Diagnosis through FMCW Radar and Machine Learning Techniques

Abstract

Respiratory diseases require early diagnosis and continuous monitoring, but existing methods involve risky physical contact. This study proposes a new system that uses FMCW radar and machine learning to monitor breathing without contact. FMCW radar can detect respiratory movements in real-time, while machine learning can classify respiratory waveforms. This study evaluates the system with cross-validation Shuffle Split, K-fold, and Stratified K-fold. The results show that Random Forest has the highest accuracy of 94.6% and Naïve Bayes has the shortest time of 0.055 seconds. Shuffle Split performs best overall. This study shows the feasibility and potential of the system for the detection, response, and tracking of respiratory diseases in emergencies.

Keywords

• Classification
• FMCW Radar
• Machine Learning
• Respiratory Detection