Determination of optimal auscultation point on the chest walls for personal identification using heart sounds

This paper propose an optimal auscultation point of heart sounds for personal iden- ti¯cation using heart sounds. The four typical auscultation points were aortic, pulmonary, Lower Left Sternal Border (LLSB) and mitral. These four views of heart sounds have been recorded for each participant who seated in a relaxed position. Several experiments were conducted using 2,314 cycles from 30 participants to ¯nd the highest recognition rate among the four locations of heart sounds. The system is developed by a combination of data analysis methods Mel Frequency Cepstrum Coe±cients (MFCC) and Gaussian Mixture Model (GMM). Preliminary results from the experiments showed the highest recognition is at location tricuspid with 94.35% when the GMM has 1 state and 32 mixtures. So, it can be concluded that the values of heart sounds at LLSB point is better than other point.It is also might be due to right ventricle walls are thinner than the left ventricle walls which has to pump for longer distances and greater cross-pressures.