A morphology alignment method for resampled heartbeat signals

A morphology alignment method for digital heartbeat signals of a person is proposed. In this method, a heartbeat is delineated into stable and flexible segments which are resampled at two different rates and then concatenated again. A resampling rate is decided based on the analysis of spatial error introduced by the interpolation method. Baseline shift of the resampled heartbeat is removed by amplitude normalization. The alignment of resampled heartbeat by this piecewise-uniform resampling method is then compared with those by uniform and non-uniform resampling methods. Four different morphological features are extracted from resampled heartbeats by each of three methods. Improvement of alignment is evaluated by to two metrics known as morphology alignment score and correlation coefficient. The method is tested on ECG signals obtained from two publicly available databases having different sampling frequencies. Statistical analyses suggest that the piecewise-uniform resampling method improves the alignment of morphology significantly. Computational time for the alignment is linear to the number of samples in a heartbeat and hence the method should be efficient enough for different practical applications using morphological features for automatic classification and decision making.

► We propose a morphology alignment method for digital ECG signal using a piecewise-uniform resampling technique.
► The method is tested on regular heartbeat signals of a large number of persons having different sampling frequencies.
► It has been found that the piecewise-uniform resampling method improves the alignment of morphology significantly.
► The method is computationally efficient to be useful in different real applications (e.g. biomedical engineering, biometrics, and emotion detection) using morphological features for automatic classification and decision making.