The Research Of Diagnosis Of Heart Disease Method Of Heart Sound Signal

Cardiovascular disease poses a serious threat to human's health, its disease mortality is the highest in all the kinds of disease. Heart sound signal is one of human important physiological signal, it contains physiological pathology information of every part of the heart. It is very important to collect heart sound signal and analysis it for improving cardiovascular disease diagnosis ability and diagnostic rate. Heart sound signal is a kind of non-stationary, time-varying, weak and complex signal, so feature extraction, classification and identification are research problems for the heart sound signal at present.In this paper, the heart sound signal collection, de-noising, feature extraction, classification and recognition of heart sound signal are analyzed. The main research contents are as follows:1. A simple, effective heart sounds signal acquisition device is made in this paper, which is combined with collection procedures of LabVIEW8.6 virtual instrument development platform to realize the collection of the heart sound signals. In the acquisition process, we can boardcast the real-time waveform of heart sound signals to improve the efficiency of acquisition.2. It is susceptible to other interfering noise in heart sound signal collection process. In order to remove the influence of the disturbed noises, this paper adopted wavelet transform method, designed heart sound signal de-noising program in the LabVIEW8.6. We can see this method has good de-noising effects through a comparison of before and after de-noising from the output waveform.3. Time series analysis method is introduced to heart sound signal analysis. Build time series autoregressive (AR) model of heart sound signal using the AR power spectrum and bispectrum estimation method, adopting LabVIEW8.6 to construct heart sounds signal power graph and 3D spectrums graph. We can see the difference obviously between the two through the analysis of comparison of the normal and abnormal heart sound signal power graph and 3D spectrums graph. To analyze the energy and phase information of sound signal which is in our research using power spectrum and bispectrum. to classify and recognize heart sound signal.Through this several aspects of the study, we can see the normal and abnormal heart sound signal exist significant differences. In the analysis of the power spectral curve, we can find that the energy of normal heart sounds signal is lower than abnormal one, it mainly concentrates in the low frequency, and its curve does not exist spectrum peak but abnormal ones do.The normal and abnormal heart sound signal is classified using distribution characteristics as characteristic information which is existing in power spectral curve. And in 3D spectrum diagram, abnormal heart sounds signal does not exist spectrum peak, its top is diverged, but normal signal exist obvious spectrum peak, its tops do not spread, its energy is low relatively. The model parameters of AR model are used as eigenvector to distinguish normal and abnormal heart sounds signal through calculation of mahalanobis distance, it achieved a good effect. This research can provide reference for the doctor and improve the effect of patient diagnosis.