Separating respiration artifact in microwave doppler radar heart monitoring by Independent Component Analysis

With a microwave radar, chest wall movements originating from cardiac and respiratory activity can be recorded non-contactly. A major challenge is how to separate the desired low-amplitude cardiac signal from large-amplitude artifacts, such as respiration. Commonly, the separation is performed with a narrow band-pass filter. This causes the signal edges to be rounded, which complicates the accurate timing estimation in the heart rate variability (HRV) analysis. In addition, the harmonics of the respiration signal might fall into the same frequency spectrum as the cardiac signal. In this study, we recorded data with two radars and studied signal separation using a complex-valued Independent Component Analysis (ICA). After ICA, the respiratory signal is greatly attenuated, although still present, in two of the independent components (ICs). However, respiration harmonics are reduced as well, and thus, the residual respiratory signal can be removed by filtering.