Dual-Stage SSA for Robust Heart Rate Estimation Using Millimeter Wave Radar

Abstract

In frequency-modulated continuous-wave (FMCW) radar-based vital sign monitoring, weak cardiac signals are often overwhelmed by strong respiration and clutter, degrading heart rate estimation accuracy. To address this, we propose a dual-stage singular spectrum analysis (SSA) method for robust heart rate extraction. Starting from MTI-filtered chirp data, a high-SNR phase signal is obtained via phase unwrapping and sliding-window filtering. The first SSA stage separates the signal into respiration-dominant and cardiac-mixed components. A residual phase signal is then formed to suppress respiration while preserving cardiac features, followed by a second SSA stage to further enhance heart-related components. Heart rate is finally estimated via fast Fourier transform (FFT) and peak detection. Experiments show that the proposed method effectively suppresses respiratory interference across various resting postures, significantly improving accuracy and robustness.