Real-Time Estimation of Respiratory Rate Using Contactless FMCW Radar with Adaptive Filtering for Various Breathing Patterns

Respiratory rate estimation is essential for health monitoring, especially for monitoring the health of the elderly. Traditional contact-based methods can be intrusive and uncomfortable. This paper proposes a contactless approach using $\mathbf{6 0}$ GHz Frequency-Modulated Continuous Wave (FMCW) radar with an adaptive filtering technique and signal quality index (SQI). The FMCW radar captures minute chest movements associated with breathing, allowing for remote and non-intrusive monitoring. Adaptive filtering narrows the band-pass filter’s bandwidth to mitigate motion artifacts and noise, improving accuracy and robustness. We evaluated the proposed method on 31 subjects with varying breathing patterns (bradypnea, normal, and tachypnea) and compared it to a contact-based ECG reference sensor. The proposed method yields mean absolute error (MAE) of $\mathbf{1. 3 9}$ and root mean square error (RMSE) of 1.55, outperforming baseline method from previous study. A real-time experiment integrating the radar sensor with a Raspberry Pi 4 enabled respiratory rate estimation every second, highlighting the potential for lightweight and portable respiratory rate monitoring tool.