Sensor fusion of capacitively coupled ECG and continuous-wave doppler radar for improved unobtrusive heart rate measurements

Unobtrusive measurement of vital signs is a research topic of increasing interest, both due to the advances in different technologies and the need for a more patient-centered healthcare. A limitation of such measurements is their sensitivity to motion artifacts, which challenges their use in daily life activities. In this paper, a novel system for the simultaneous acquisition of multichannel capacitively coupled ECG (ccECG) and radar signals is presented, which allows for respiration rate and heart rate (HR) extractions. Furthermore, a combination of signal quality and confidence indicators is proposed, which serves as a first step towards sensor fusion for an increased accuracy and time coverage of HR monitoring. After applying the proposed quality estimation algorithms, sensor fusion methods were performed including quality-based ccECG channel selection, confidence-based radar HR source selection, and Bayesian fusion. Both the system and the proposed algorithms were tested in a total of 125 min of measurements obtained from five different volunteers performing normal office work. An improvement in time coverage (defined as the time in which the HR error is lower than 3 bpm) from 2% to 32.5% was obtained for radar signals with the proposed confidence-based algorithm; the same metric for ccECG increased from 52.2% to 82.1% when automatically selecting the periods with the availability of good quality ccECG signals, while achieving 35.7% coverage in the remaining periods, by using one of the proposed Bayesian fusion methods; errors in each case were reduced accordingly, and a clear division of two sets of signals with different (increased) coverage levels was achieved.