Model-based ECG fiducial points extraction using a modified extended Kalman filter structure

This paper presents an efficient algorithm based on a nonlinear dynamical model for the precise extraction of the characteristic points of electrocardiogram (ECG), which facilitates the HRV analysis. Determining the precise position of the waveforms of an ECG signal is complicated due to the varying amplitudes of its waveforms, the ambiguous and changing form of the complex and morphological variations with unknown sources of drift. A model-based approach handles these complications; therefore a method based on the usage of this concept in an extended Kalman filter structure has been developed. The fiducial points are detected using both the parameters of Gaussian-functions of the model, and the state variable estimates. Several MIT-BIH ECG records are used for performance evaluation. Results show that the proposed method has an average sensitivity of 100% and a specificity of 99.93%. Simulation results illustrate that the method can contribute to and enhance the clinical ECG points extractions performance and exact tachogram representation.