Towards a time-feature independent phonocardiogram segmentation

Delimitation and classification of each heart sound is a rather difficult task. Elevated heart rates, as found in pediatrics and in some adults as well, influence some of the most reliable features used by existing methods. Furthermore, in real life scenarios, cardiologists will not have the time to acquire the signal´s length required by some of the existing algorithms, which make us think that different approaches ought to be pursued. This paper presents the work on heart sound segmentation using structural and energy based features. It is an attempt to not rely on features considered crucial to most existing approaches. Yet, it achieves a high sensitivity and specificity comparable to some literature.