A Multi Rate Marginalized Particle Extended Kalman Filter for P and T Wave Segmentation in ECG Signals

The marginalized particle extended Kalman filter (MP-EKF) has been known as an effective model-based nonlinear Bayesian framework in the field of electrocardiogram (ECG) signal denoising. In this paper, we reveal another potential capability of an MP-EKF and propose a multirate MP-EKF based framework for P- and T-wave segmentation in ECG signals. The proposed multirate implementation of MP-EKF leads to better estimation of states and avoids unwanted errors in estimation procedure. The behavior of particles in the multirate MP-EKF is controlled by a novel particle weighting strategy that helps the particles adapt themselves with respect to ECG signal trajectory. After ECG filtering, a novel morphology-based algorithm uses the estimates of a multirate MP-EKF to determine the Pand T-wave fiducial points. This algorithm is a combination of well-known morphological operators such as "opening," closing, "top-hat," and "bottom-hat" transforms. The segme ntation performance of the proposed algorithm was evaluated on QT database and it showed promising results in comparison to other Bayesian frameworks such as partially collapsed Gibbs sampler and extended Kalman filter.