Research And Application Of Multichannel Heart Sound

As a biological feature,heart sound has great practical value.In the traditional medical field,heart sound has become a common means of diagnosing heart disease and evaluating heart function.In the emerging field of biometrics,heart sound has become a new research hotspot.However,the heart sound signals collected from different auscultation positions have similarities and certain differences.Fully studying and making good use of these commonalities and differences will help to improve the application value of heart sound in various fields.In this paper,the multi-channel heart sound hardware acquisition module,the data acquisition and storage module based on Lab VIEW,and the algorithm analysis module based on MATLAB are combined to conduct in-depth and detailed research on multi-position heart sound characteristics.The heart sound hardware acquisition module integrates the self-made 4-channel heart sound sensor,LM386 power amplifier circuit,5V DC power supply and the hardware acquisition board card USB-4432 of NI company,which synchronously records the heart sound signals from the four auscultation areas of mitral valve,tricuspid valve,pulmonary valve and aortic valve.The data acquisition and storage module is developed and designed based on the Lab VIEW platform,which realizes the real-time display of four-channel heart sound signal waveform,and saves the personal information and heart sound data of the collector.The analysis module based on MATLAB reads the heart sound data collected by the Lab VIEW platform.Firstly,the four-channel heart sound is denoised and segmented.The db6 wavelet threshold denoising method is used to suppress the low-frequency and high-frequency noise of heart sound,and the HSMM segmentation method based on Logistic regression is used to segment heart sound accurately.Then extract the heart rate(HR),diastolic and systolic time ratio(D/S),the first heart sound and second heart sound amplitude ratio(S1/S2),these three kinds of important indicators reflect cardiac reserve capacity,a multi-position fusion cardiac function evaluation method based on Dempster-Shafer evidence theory was proposed,and a cardiovascular health evaluation report was generated,which is more reliable and reasonable than the single-channel heart sound evaluation.The method of linear predictive coding(LPC)was used to extract the formant frequencies of S1 and S2 as feature vectors,a method of heart sound biometric identification based on formant frequency and Euclidean distance was proposed,and the influence of different auscultation positions on the recognition effect is studied.The experimental results show that the recognition rates of different auscultation positions are slightly different: The recognition rate of pulmonary valve was 92.5%,that of aortic valve was 90%,that of mitral valve and tricuspid valve was 87.5% and 85%,respectively,and the heart sound signal of pulmonary valve was basically not related to the gender of the subjects,this provides a practical reference for heart sound acquisition and recognition.