Extraction of vibration feature using purpose designed wavelets for heart diagnosis

The heart sounds with systolic murmurs can be characterized and visualized as a result of free and forced vibrations in human hearts. A joint time frequency distribution with oscillating windows is proposed to extract the signals from heart sounds. Two types of mother wavelets are designed as matching templates based on the nature of murmurs. The time-frequency maps of the analyzed phonocardiographic (PCG) signal are compared to show which type of templates matches better with the systolic murmurs. Practical applications are addressed for two types of cardiac abnormalities, aortic stenosis and mitral regurgitation. It has been demonstrated that by means of using different oscillating window functions, different characteristics of the heart sounds associated to the heart disease can be enhanced in the time-frequency domain analysis.