CRACS Seminar: Exploring the Stationary Wavelet Transform for the detection and identification of the S1 and S2 Heart Sounds

13 March 2013

Nuno Marques, CRACS new collaborator is starting his participation in the CRACS team by giving a seminar! He is an M.Sc. student of our department and will give a seminar about his work on identification of cardiac sounds, next Wednesday, Mar 13th, at 2pm, in room S2 of DCC.

In order to do heart sound segmentation, i.e. detect and identify S1 and S2 heart sounds from cardiac auscultation signals, most approaches are based on the Shannon energy. This kind of approach requires a noise reduction phase
followed by the computation of the Shannon energy of the signal, and the detection of its peaks. We propose a new method that uses the Stationary Wavelet Transform to segment the signal and hierarchical clustering to distinguish the S1 and S2 heart sound from noise. This approach was tested in the Dataset B of the Classifying Heart Sounds PASCAL Challenge and achieved significantly better results than the winning approach of this contest.