An Optimized Signal Quality Assessment Method for Noncontact Capacitive ECG

Noncontact capacitive electrocardiogram (cECG) is gaining recognition in cardiovascular disease monitoring for its comfort and noninvasiveness. Compared to the conventional electrocardiogram (ECG), cECG signal quality is prone to degradation in practical applications due to motion artifacts and power line interference (PLI). This study proposed an optimized signal quality assessment method to identify and remove low-quality cECG signals. First, the human body-electrode interface is modeled to analyze the generation mechanism and influence of cECG motion artifacts and PLI. Then, distinct signal quality indices (SQIs) are proposed to target the characteristics of these interferences. Moreover, optimized cECG features and previously proposed ECG features were combined as multifeatures and presented to XGBoost for binary classification training. Finally, Shapley additive explanations (SHAPs) were utilized for feature optimization to reduce redundant information. Validation on a labeled noncontact cECG database yields an impressive binary classification accuracy of 98.786%, an ${F}1$ -score of 98.845%, and a kappa of 97.567%. Moreover, its performance on a subject-independent validation set is also excellent, with an accuracy of 99.130%, an ${F}1$ -score of 96.937%, and a kappa of 96.430%. The optimized multifeatures also demonstrate favorable performance in a triple classification model. The experimental results show that our method offers a precise and convenient solution for cECG signal quality assessment.