The Indirect Estimation of Breathing Rate through Wearables: Experimental Study and Uncertainty Analysis through Monte Carlo Simulation

Breathing Rate (BR) is a fundamental physiological parameter and wearable sensors can indirectly estimate it through the measurement of electrocardiogram (ECG). Indeed, they are widely employed in several application fields thanks to their multiple advantages, such as user- friendliness, availability in different quality and cost segments, and capability to acquire multidomain physiological signals. This study aims at applying an approach based on respiratory sinus arrhythmia to the ECG signals acquired by a cardiac belt (Zephyr BioHarness 3.0) and a smartwatch (Samsung Galaxy Watch3), evaluating the measurement accuracy as well as performing a Monte Carlo simulation to analyze the uncertainty propagation along the measurement chain, from the wearable sensors to the estimated BR value. The results show that both the wearable sensors provide an accurate estimation of BR (almost null bias), with good precision (standard deviation of residuals: 3 bpm for both sensors), and moderate-high correlation with reference values (Pearson's correlation coefficient: 0.77 for Zephyr BioHarness 3.0 and 0.63 for Samsung Galaxy Watch3). Considering an uncertainty of ±1 bpm and ±2 bpm on heart rate for Zephyr BioHarness 3.0 and Samsung Galaxy Watch3, respectively, the Monte Carlo simulation provided expanded uncertainty values on the estimated BR of ±6 bpm and ±8 bpm, respectively, evidencing a relevant impact of physiological variability.