Research On The Synthetic Model And Analysis Features Of The Heart Sound Signal

The heart sound signal is one of the most important physiological signals for human, itsproduction process and composition are very complex. In most situations, we can only hear thesetwo components of the first and second heart sounds in the cardiac auscultation. Their changing isoften in relationship with organic heart disease and can also reflect that different heart values ormyocardial have pathological changes. Before, many scholars consider that the first heart sound andthe second heart sound are caused by the collision in the process of value opening and closing.These vibration sources are all transient non-stationary signals. Therefore, the establishment of theheart sound signal generation is one of the most effective tools for learning and auscultation of theheart.In this thesis, According to the the generation mechanism of the heart sound signal, Acomputational model of the cardiovascular system is described to model the first heart sound. Themodel provides a framework for studying quantitative physiological models of heart soundgeneration. The first heart sound generation process is built based on the hemodynamic variables.Parameters of the cardiovascular model can be adjusted by users to produced different sound signalsunder various pathological conditions. But,the effect of the heart sound auscultation is also to beimproved further. Then, A non-linear transient chirp signal model method is described for thedecomposition and synthesis of the aortic(A2) and pulmonary valve(P2) components of the secondheart sound(S2). The method is based on the time-frequency representation, which can be used toestimate and reconstruct the instantaneous phase and amplitude functions of each component.In orderto verify the accuracy of the method, a simulated S2with A2and P2components having differentoverlapping time intervals which is betwwen5and30ms was synthesized. The simulation resultsshow that the approach is very effective for extracting A2and P2components. Finally, the heartsound signal was generated by synthesizing the first and second heart sound.To evaluate the accuracy of the proposed heart sound signal models, A wavelet decompositionand normalized Shannon energy envelope extraction methods are used for analysis of the simulatedand standard first heart sound and the second heart sound. The simulation results show that thesimulated heart sound components are similar with the actual heart sound signals very highly. A studying on the classification and identification of the heart sound signals becomes one ofthe focus of the current research gradually. A emergence or changing of the heart sounds andmurmurs can usually reflect different symptoms of the heart diseases. This thesis is based on thenormal heart sound signals and several abnormal heart sounds. Power spectrum estimation andnormalized Shannon energy algorithms are adapted to obtain some typical characteristics ofparameters in the time and frequency domain. Thus, different heart sound signals can be separatedeffectively from different aspects, which provide a favorable basis for the diagnosis of heartdiseases.