Heart rate estimation algorithm for radar signals based on wavelet threshold noise reduction and Resnet50

Non-contact heart rate monitoring techniques based on millimeter wave radar have advantages in medical scenarios. However, the accuracy of existing heart rate estimation methods is still limited by DC (Direct Current) offset, respiratory harmonics, and environmental noise interference. And the observation time needed for heart rate estimation is still long. To address these problems, this paper proposes a heart rate estimation algorithm for radar signals based on wavelet threshold noise reduction and Resnet50. In the phase preprocessing stage, phase difference, interpolation smoothing, and wavelet threshold denoising are performed on the phase signal. In the heart rate estimation stage, Resnet50 (Residual Network) accepts the phase signal for heart rate estimation. The wavelet threshold denoising provides Resnet50 with a noise-reduced phase signal, and Resnet50 may solve the problem of respiratory harmonic interference, which is challenging for some traditional signal processing methods. Experimental results show that the way in this paper improves the accuracy of heart rate estimation while reducing the required observation time.