ECG Segmentation by Adaptive Rational Transform

We propose a novel method to delineate and segment ECG heartbeats, i.e. to provide model curves for the P, QRS, and T waves, and to extract the fiducial points of them. The main idea of our method is to apply an adaptive transformation by means of rational functions to model the heartbeats and their waveforms. We suggest to represent the heartbeats with a linear combination of rational functions that are selected adaptively to the ECG signals through a non-linear optimization. This leads to a simple, yet morphologically accurate description of the heartbeats, and results a direct segmentation of them. Then, we derived the fiducial points based on the analytical model curves extracted from the rational representation. Multiple geometric concepts and their combination is discussed to this order. The evaluations were performed on the QT Database, and the results are compared to the previous ones, proving the efficiency of our method.