Dual-Radar Integration for Vital Signs Acquisition under Heavy Body Movement using Machine Learning

The exploration of non-contact vital sign detection through Doppler Radar systems has gained significant attention and research within the healthcare domain over the past decade. While wireless vital sign detection methods are already reliable, there is a notable focus on scenarios involving stationary subjects. This article delves into a novel approach, concentrating on capturing respiratory signals under heavy body movement. The proposed system uses Deep Learning (DL) algorithms for signal reconstruction and employs two 5.8 GHz carrier radars to validate the efficacy of body motion compensation. The primary objective is to verify whether the use of two radars offers advantages over a single radar when subjects are walking on a treadmill. This success encourages further exploration in environments characterized by continuous heavy body movement. The study concludes with a detailed analysis, indicating that employing two radars does not provide a discernible advantage over using a single radar. Even in situations where radar-acquired data exhibits minimal respiratory signals, Deep Learning demonstrates the capability to accurately reconstruct vital signs. The errors attained in this paper demonstrate the potential of DL for respiratory signal reconstruction acquired using radar, attaining errors around 1 Respiration per Minute (RPM) across the networks studied, it also allowed to prove that the use of more data, that being the use of two radars, in this application, only shows benefits in one network.