Methods On Noise Reduction And Segmentation Of Heart Sound Signal

Heart auscultation is an important method for diagnosing heart valve diseases. The auscultation, however, is always effected by doctors' subjectivity and cardiac auscultatory skills. So it's necessary to take advantage of the computer to assist the diagnosis. Phonocardiogram (PCG) can provide valuable information about heart valve function and cardiac hemodynamics, and assist doctors in detecting and diagnosing various heart diseases.Heart sounds are no stationary, transient signals. They are easily contaminated by noise. This paper proposes a new method in joint time, frequency, medicine domains to reduce noise. Heart sound recording is decomposed into a series of time-frequency atoms, characterized by amplitude, time, frequency, duration by the modified matching pursuit method. From the features of heart sound in medical domain, it is found that heart sounds behave as clusters in time-frequency domains in nature. A clustering method is thus employed to distinguish the atoms for heart sounds. Meanwhile, the atoms for noise, which are out of clusters, are rejected. Experiments on various heart disease's cases are made, and the similarity between the original signal and the reconstructed signal exceed 90%, which shows the validity of this de-noise method.Heart sound is generated by the combined results of cardiac valves activities and related vibrations. The parameters of duration, amplitude, frequency reflect the working conditions of cardiovascular. Segmentation of heart sound is an important step for automatic analysis by computer. In this paper, we propose a method to segment the first heart sound and the second heart sound in time-frequency domain according to heart sound's characteristics. Firstly, a time-frequency plane is gained by time-frequency analysis, and then time displacement and frequency information of each component are determined. We can make certain different regions by searching and determining the previous information and thus to establish the scope of the corresponding components of heart sound. We have made experiments on the heart sound signal from many heart disease's cases. The results of segments for S1 and S2 are satisfactory. The method proposed in this paper can provide an important basis for the next step of analysis and diagnosis of heart sounds.