Beat to beat classification of long electrocardiograms using entropies and hierarchical clustering.

This paper introduces an entropy based method for beat to beat classification of long electrocardiograms (ECGs). A state vector is reconstructed using Taken's delay coordinates method and Shannon entropies are calculated for each beat to form feature vectors. Hierarchical clustering is applied to these vectors to classify the beats. The algorithm was used for detection of atrial premature beats and ventricular premature beats in long electrocardiograms.