Abstract: In order to solve the problem of low signal-to-noise ratio for radar echo signals in a sheltered environment and the masking of human heartbeat signal by respiratory harmonics and other clutter, a noise reduction method based on improved singular value decomposition (ISVD) and a vital signs extraction method based on improved time-varying filtering empirical mode decomposition (ITVFEMD) is proposed. First, the radar echo signals are preprocessed using the improved singular value decomposition algorithm. Subsequently, the preprocessed signals undergo parameter-optimized ITVFEMD to decompose the signal into intrinsic mode functions (IMFs). Appropriate IMFs are then selected to reconstruct the respiratory and heartbeat signals. The reconstructed signals are analyzed via Fast Fourier Transform (FFT) to estimate the corresponding respiratory and heartbeat frequencies. Finally, the estimated frequencies are compared with ground truth data obtained from a pulse oximeter for consistency analysis. Experimental results show that the proposed method achieves estimation accuracies of over 95% for heart rate and 98% for respiratory rate.

Key words: sheltered environment, ITVFEMD, vital signs extraction