A Long-Range Vital Signs Sensing Framework Using Massive Millimeter-Wave Channels

Abstract

Long-range vital signs sensing using millimeter-wave (mmWave) radar faces two challenges: one is the severe path loss in mmWave bands, which results in reflected signals from the subject being too weak and drowning in noise. The other is that the subtle movements of breathing and heartbeat make the vital signals too faint to be obtained. Inspired by channel diversity, a long-range vital signs sensing framework using massive channels is proposed to tackle these challenges. First, considering the width of the torso, a vital bins selection method based on diversity combining and correlation analysis (DC-CA) is proposed for identifying the range bins with vital signals. Second, an improved independent vector analysis based on principal component analysis (PCA-IVA) is put forward to enhance the vital signals from massive channels by leveraging their quasiperiodicity and correlation. Third, quality factor (QF) variation for parameter optimization and dispersion entropy (DE) for vital components selection are introduced into multivariate variational mode decomposition (MVMD) as the MVMD-QF-DE method to separate respiratory and heartbeat signals with high quality. The experimental results demonstrate that the proposed framework could accurately sense vital signs even when the reflected signals and vital signals are drowned out by noise, thereby increasing the sensing range to 25 m. The sensing accuracy is significantly improved compared to existing methods, achieving respiratory and heartbeat rate accuracies of 97.25% and 98.84%, respectively.