A new algorithm for removing random motion noise of human body based on total variation

Abstract

Millimeter-wave radar sensors have garnered significant attention in non-contact monitoring for detecting human vital signs, particularly respiration and heart rate. These sensors offer advantages such as compact size, lightweight design, and versatility in sensing across diverse scenarios. Among them, frequency-modulated continuous wave (FMCW) radar demonstrates considerable potential for vital sign monitoring. However, a critical challenge persists, human random motion noise, which spans the entire frequency domain, significantly interferes with accurate heartbeat signal detection. To address this issue, this paper proposes a total variation model-based approach. This method involves reconstructing the thoracic signal matrix, converting it into a grayscale image, and obtaining the sparse characteristics of noise and the underlying image structure based on the second-order gradient information in different directions of the image. By introducing the Lp pseudo-norm, the second-order regularization constraint terms in the horizontal and vertical directions are designed respectively. Meanwhile, a global sparse constraint term was designed based on the prior characteristics of the noise. Then, the noise and the sparse characteristics of the underlying image structure are utilized for denoising, and finally, the thoracic cavity signal is reverse-reconstructed. This framework effectively suppresses random body motion noise while preserving vital sign information. Experimental results demonstrate a notable improvement in signal-to-noise ratio (SNR), along with enhanced measurement accuracy for both respiratory rate and heart rate. The findings of this study not only advance the theoretical framework for non-contact vital sign monitoring but also underscore the practical utility of FMCW radar in real-world applications.