Indirect Estimation of Breathing Rate through Wearable Devices

Breathing rate (BR) represents one of the most important physiological parameters to be measured in clinical environment, being linked to multiple stressors and hence describing the global well-being state of a subject. During the last years, both contact and non-contact measurement methods for the assessment of BR have been proposed, particularly for ambulatory remote use. Given the wide spreading of wearable devices, several indirect estimation methods from heart rate (HR) data series have been proposed, avoiding standard procedures that can be often invasive or intrusive. This study aims at estimating the measurement accuracy and precision of BR when indirectly estimated from the HR series gathered by wearable devices (i.e. wrist-worn and chest-strap sensors). Volunteer subjects were tested both in natural breathing conditions and making them inhaling and exhaling at determined frequencies. The estimated BR values were compared with the reference ones and the accuracy and the precision of the measurement were evaluated through standard techniques (i.e. analysis of deviations, evaluation of agreement and correlation analysis). The study shows that better results are obtained with chest-strap sensors thanks to the higher accuracy of HR data: mean deviation of -0.24 bpm and -0.40 bpm for BioHarness 3.0 and Polar H10, respectively, with respect to the mean deviation of 2.89 bpm reported for Garmin Venu Sq for accuracy; standard deviation of 1.36 bpm and 2.92 bpm for BioHarness 3.0 and Polar H10, respectively, with respect to the mean deviation of 4.74 bpm reported for Garmin Venu Sq for precision. It is also noted that better BR accuracy and precision can be obtained by dedicated signal processing.