Multi-Subject Heart Rate Estimation and Real-Time Tracking Using a mmWave Radar and Trace Carving Algorithm
In this paper, accurate real-time heart rate tracking for multiple subjects is explored. A signal processing scheme implemented using a 60GHz, small form factor FMCW radar and compact processing unit is presented. Improvements in target discrimination and heartbeat signal-to-noise ratio (SNR) are demonstrated utilizing the MIMO processing capabilities of the radar. For target discrimination, direction of arrival algorithms are used to automatically detect the number of subjects in a given scene, as well as their precise angular location relative to the radar. Subsequently, digital beamforming of the channels is used to maximize the SNR of the relatively weak heartbeat signal of each subject. Finally, a real-time frequency trace carving algorithm is used to accurately track the heartbeat signal of each subject over time. The presented method enables real-time tracking using a Jetson Nano, a compact microcomputer, through a one-second update window and offers an average mean absolute error of better than 5bpm across a wide range of scenarios.