Radar-Based Vital Signal Classification: Detecting Random Body Movement and Inter-Modulation Products with Temporal Convolutional Networks

This paper presents a novel approach using a 60-GHz Frequency Modulated Continuous Wave (FMCW) radar system for vital signal classification, particularly focusing on identifying random body movements (RBMs) and inter-modulation products (IMPs). Leveraging Temporal Convolutional Networks (TCNs), our methodology achieves promising results in distinguishing between various signal types. Through rigorous testing on data from seven healthy individuals, we achieved an accuracy of 98.75% for distinguishing RBM signals from high-quality data (without IMPs and RBMs), and 82.25% for distinguishing high-quality data from RBMs and IMPs on the test dataset. Notably, we address the critical issue of discerning between good and bad data, serving as an intelligent filter to enhance subsequent estimation algorithms. This filtering mechanism significantly reduces the risk of inaccurate estimates, thereby improving the reliability and efficacy of vital signal estimation systems.