Noise Reduction in Capacitive ECG Measurements by Dedicated Modelling and Kalman Filtering

Electrocardiography (ECG) measurements have been extensively used for monitoring and diagnosing cardiovascular diseases (CVDs). Traditional ECG measured with wet electrodes may be unsuitable for long-term monitoring due to the use of gels that may cause skin problems. A capacitive electrode can measure an ECG though an isolating layer and is therefore a good alternative to wet electrode for long-term ambulatory monitoring. However, current application of capacitive ECG (cECG) focuses mainly on heart-rate analysis due to its high sensitivity to noise. In the present study, by using a dynamic ECG model, we propose an extended Kalman filter (EKF) for cECG noise removal. For evaluation, cECG signals were recorded from 8 healthy subjects and processed by the proposed method. The denoised cECG were then compared with the traditional ECG recorded simultaneously using gel electrodes using the Pearson's correlation coefficient (CC) as performance metric. To assess the quality of the detailed ECG waveform, such as P and T waves, CCs were also calculated after removing the QRS complex from both signals. Our results show EKF to produce promising results in both CCs, outperforming the state-of-the-art methods. Particularly, the observed high CC in the detailed ECG waves, i.e., 0.81 on average, indicates the feasibility of morphological analysis using cECG, enabling possible clinical application of cECG.