Development Of Analysis And Recognition System On Heart Sounds

Heart sound signal is one of the most primary physiological signals. Detection of heart sounds is an important method with judging the state of heart which has its own advantages that ECG can not replace. Close of heart valves has a voice, the value of whose main frequency will increase if artificial heart valves is calcified. So degradation assessment of artificial heart valves can be detected with the spectrum of close voice. Amplitude of the first heart sound is the normal measurement of cardiac contractility, therefore, the heart sound can be used to evaluate cardiac contractility. Phonocardiograms (PCGs) visualize cardiac auscultation, and improve the diagnosis level of cardiovascular diseases. It is convenient to analyze the heart sound and extract the characteristic parameter for doctors to obtain basic information of heart and make more exact diagnosis.The PCG signal is non-stationary. In order to get a comprehensive understanding of the characteristic for the heart sound, it is necessary to analyze heart sound in time, frequency and time-frequency domain. Due to the interference, the heart sound signal is noised so that the preprocessing is necessary to remove the noises. After the preprocessing, the quality of signal is improved and can be used for further analysis.Based on the theories of Hilbert Transform, Short Time Fourier Transform (STFT), Wavelet analysis and Huang Transform, this paper makes a deeper research on the analysis of heart sound and the design for heart sound analysis system. The main contents are listed below:(1) To remove the low frequency noise in heart sound, a new method based on Huang Transform is proposed in this paper. It may decompose signal into series of intrinsic mode function (IMF) with empirical mode decomposition (EMD) and the low frequency noise can be isolated because of its different of frequency in this process. The experiment results show that this method can remove the low frequency noise more effectively than the wavelet filter.(2)To make characteristic in time domain more accurate, Hilbert Transform is used to extract the envelope of heart sound. The signal is transformed to the analytical signal by Hilbert Transform and the absolute value of that is the envelope of heart sound. Compared with normalized average shannon energy, this method not only can get more accurate result but also can extract envelope when heart murmurs occur.(3) In order to get quantitative analysis for the result of time-frequency, the traditional methods on time-frequency analysis are improved. Because the result is three dimensional data, it is difficult to get characteristic quantitatively. In order to get two dimensional distribution on time-frequency, the result is divided into several parts in time domain. In every part, frequency which has the biggest amplitude is the adaptive frequency for this part. With this method, the frequency of every time can be obtained.(4) To operate heart sound analysis system easily, the system is developed by Matlab GUI which is a graphical interface development environment provided by Matlab. The interface is designed with control modules and the function is implemented with callback functions.Finally, we make a conclusion and propose the future research directons in this field.