Classification of atrial fibrillation episodes using short ECG segments

This paper presents a method to classify different subtypes of atrial fibrillation episodes by analyzing short segments of electrocardiograms. We will process surface ECGs segments by time-frequency transforms to extract relevant features that will be used as input to a neural network classifier. As atrial fibrillation presents a progressive nature, this method can be a very useful tool in order to differentiate the progress of the arrythmia in each patient.