Improving Signal Quality in Non-Contact Electrocardiography: Novel Strategy for Motion Artifact Reduction

Capacitive electrocardiography (cECG) enables non-contact heart rate monitoring through clothing, but motion artifacts remain a critical limitation for practical applications. We present a novel motion artifact removal method using non-contact floating electrodes as noise references combined with multi-reference Normalized Least Mean Squares (NLMS) adaptive filtering. The floating electrodes, positioned without skin contact, couple primarily to ambient 50 Hz mains interference, which becomes amplitude-modulated during motion due to changes in electrode-body capacitance. Six reference signals are derived from this noise electrode: band-pass-filtered signal and its derivative (capturing baseline-type artifacts), envelope and its derivative (capturing amplitude modulation patterns), and envelope asymmetry and its derivative (capturing non-linear electrode response during motion). The NLMS algorithm adaptively combines these references to estimate and remove motion artifacts while preserving QRS morphology through low-pass filtering of the correction signal. A hysteresis-based motion detector with minimum duration constraints enables selective application of artifact removal only during motion periods, leaving rest-period ECG unmodified. We present this as a proof-of-concept validation of a novel reference-electrode architecture for motion artifact suppression in non-contact ECG. The method was validated on 7 subjects across 24 recording sessions using two electrode configurations in two environments with different electromagnetic interference levels. Controlled axial rotation motion was induced at three frequencies using a custom apparatus with IMU-based gamification for protocol adherence. Performance was evaluated using R-peak detection F1 score against gel surface-contact electrodes ground truth and RMS reduction in motion regions. Results demonstrate consistent improvement in R-peak detection accuracy during motion periods with substantial artifact energy reduction. The proposed method is designed to address motion artifacts regardless of their physical source, though the present validation focused on subject-induced motion.