Virtual Radar Model and Processing for All-Directional Measurements of Vital Signs

Abstract

To address the limitations of the radar horizontal field of view (HFOV), this article proposes a novel virtual frequency-modulated continuous-wave (FMCW) radar model to achieve simultaneous all-directional measurements through the integration of rotation motor and radar. However, weak vital signs are susceptible to interference from vibrations caused by the torque ripple of motor, which leads to the generation of ghost objects and phase noise interference. First, a Doppler combination filtering (DCF) method is proposed to localize the subjects and suppress interference from torque ripple vibrations, utilizing multiple Doppler frequencies associated with breathing for improved precision. Second, an adaptive convergence variational mode extraction-weight pair accumulation (ACVME-WPA) method is proposed to separate breathing and heartbeat signals from the phase signal affected by motor vibration. This method incorporates two key ideas: one involves determining the adjustment rate of the penalty value based on the iterative differences between center frequencies and the other focuses on aggregating multiple phase signals using weighted pairs to enhance performance by considering the HFOV overlap of adjacent virtual radars. Experimental results show that the proposed methods can achieve all-directional measurements of localization, respiration rate (RR), and heartbeat rate (HR) for single subjects with various breathing patterns, with HR mean absolute error (MAE) of 0.026 Hz in the breath-holding experiment, and also work well for multiperson scenarios with different angles and window lengths. In the four subjects’ experiment, the proposed method still works even when subjects are not facing the radar. For right-side orientation, the RR and HR MAEs were 0.007 and 0.025 Hz, respectively.