Multidomain Separation for Human Vital Signs Detection With FMCW Radar in Interference Environment

Noncontact human vital signs detection based on frequency modulated continuous wave (FMCW) radar exhibits significant application values in medical health, disaster relief, and public security. In this study, we propose a novel method named symplectic geometry variational mode decomposition (SG-VMD) for effectively separating phase signals of radar echoes in both time and frequency domains. In the time domain, symplectic geometry similarity transformation (SGST) is employed for coarse decomposition, which maintains the integrity of the temporal structure of the signal. Afterward, according to the orthogonality of signal components in time and frequency domains, we adopted variational mode decomposition (VMD) to finely decompose components obtained from SGST. Eventually, the second-harmonic weighted compensation selection (SHWCS) is proposed to identify the most suitable heartbeat signal components in interference environments. In addition, the optimized hyperparameters are determined through the dingo optimization algorithm (DOA), which uses a newly proposed fitness function named inverse envelope entropy (IEE). The efficacy of SG-VMD in interference environments is further demonstrated via the design of experiments scenarios presenting weak heartbeat signals and colored noise interference. The experimental results indicate that the SG-VMD can accurately detect human heart rate (HR) with a high accuracy of 99.016% in coverage interference scenarios and 85.714% in colored noise interference scenarios.