Preliminary comparison of respiratory rate estimation methods using simultaneous dual- position under-mattress millimeter-wave radar

Background Under-mattress millimeter-wave radar may provide unobtrusive respiratory sensing, but signal morphology can vary with sensor position and estimation method. This study aimed to compare frequency-domain and time-domain respiratory rate estimation using simultaneous dual-position under-mattress radar measurements. Methods Three healthy adult volunteers underwent supine instructed breathing at 10, 12, and 15 breaths/min. Two XM125 60-GHz-class radar modules were operated simultaneously beneath an 8-cm non-coil medical mattress at positions aligned with the sternal angle and xiphoid process. Each condition was repeated three times, yielding 27 paired trials and 54 radar recordings. The respiratory rate was estimated using a power spectral density method and time-domain peak detection. Mean absolute error, root mean square error, paired position differences, and exploratory harmonic characteristics of peak-detection failures were examined. Results Estimation based on power spectral density yielded a mean absolute error of 0.472 breaths/min and a root mean square error of 0.479 breaths/min, compared with 1.710 and 2.250 breaths/min for peak detection. The mean paired difference between sensor positions was +0.037 breaths/min for power spectral density estimation and +0.164 breaths/min for peak detection. Power spectral density estimates differed by no more than 0.05 breaths/min in 26 of 27 paired trials. Recordings with peak-detection errors greater than 1.0 breaths/min showed higher second- and third-harmonic ratios. Conclusion Under controlled instructed-breathing conditions, power spectral density-based estimation showed smaller deviations from the instructed breathing rates than simple peak detection, whereas the two positions generally yielded similar frequency-domain estimates. These findings are technical results relative to instructed breathing rates and do not constitute clinical validation.