Remote Vital Sign Monitoring With Reduced Random Body Swaying Motion Using Heartbeat Template and Wavelet Transform Based on Constellation Diagrams

Remote vital signs monitoring using millimeter-wave (mmWave) sensors has gained lots of attention because of their contactless, portable advantages. However, their received signals are more sensitive to random body motions (RBM) which degrades the accuracy of heart rate (HR) detection. To overcome this challenge, multi-input multi-output (MIMO) configuration can be used to reduce RBM's impact as each channel has different points of view with respect to the subject under test (SUT). Here we propose the use of a Frequency Modulated Continuous Wave (FMCW) radar from Texas Instruments (TI) at 77 GHz to collect data from its 192-channel configuration. Since vital sign information extracted using Arctangent Demodulation (AD) could be corrupted by either RBM or respiratory harmonics, a method is needed to minimize such effects. Hence, we develop an algorithm where a Heartbeat Template (HBT) is extruded based on the Constellation Diagram that shows the Quadrature signals from the target's range profile. The HBT is then used to design an adapted-wavelet for Continuous Wavelet Transform (CWT) to magnify the heartbeat signals. Under circumstances where RBM overwhelms the heartbeat signal that the HBT cannot completely reduce its effects, a spectral-based HR selection method is also developed to estimate the HR. By employing the proposed methods, we have reduced mean-error ${\mu _e}$ of HR estimation significantly from 9 bpm to less than 2 bpm.

I. Introduction

New generation of mmWave sensors with compact size have made contactless human vital signs monitoring more viable for a variety of applications [1]–[7]. Since the concept of remote vital sign detection is based on sensing physiological motion of the chest's skin, interference from RBM certainly impacts the detection accuracy. This interference remains to be the most difficult issue to address to date. Generally, RBM can be divided into two categories: 1) from the radar itself [8]–[9] and 2) from the subject under test (SUT) [10]–[21]. For the first category, [8] placed a motion sensor onto the hand-held radar to record any unwanted vibration caused by hand-shakes. Thus, the hand-shake vibration can be cancelled based on motion-sensor data. Meanwhile, [9] proposed the use of Empirical Mode Decomposition (EMD) technique to cancel out the unwanted vibrations as they have higher frequencies that can be realized from Intrinsic Mode Functions (IMFs).