An Adaptive Variational Modal Decomposition Method For Vital Signs Extraction

In this paper, a novel method, named EE-APVMD is proposed for vital signs extraction using a 60GHz FMCW radar system. Variational Modal Decomposition(VMD) is commonly used for the extraction of vital signs, but its parameters are often chosen empirically. Envelope Entropy(EE) reflects the sparse characteristics of the signal, and the periodic information contained in the Intrinsic Mode Function(IMF) is positively correlated with the sparse characteristics and negatively correlated with EE. Thus this paper takes EE of maximal energy IMF as the indicator of the minimal optimization model, and Particle Swarm Optimization(PSO) algorithm is used to find the best combination of the number of modes(k) and quadratic penalty factor(α) in VMD. Experimental data from 15 volunteers verify the inverse relationship between EE of each IMF and correlation coefficient between the selected IMF and the standard heartbeat data. Besides, compared with other methods, EE-APVMD has better performance in accuracy and correlation coefficient with standard data.