Research On Heart Sound Acquisition And FCNN Classification Algorithm Based On Bluetooth 5.0

Heart sound is one of the main physiological parameters in the diagnosis of cardiovascular diseases.Traditional auscultation of heart sounds is widely used in the early screening of heart diseases,the diagnosis results are mainly based on the subjective opinions of physicians,which are prone to misdiagnosis and missed diagnosis.This topic aims to develop a better heart sound acquisition equipment to help the establishment of heart sound database,and design an effective heart sound recognition algorithm for the classification and discrimination of normal and abnormal heart sounds,assist doctors in diagnosis,and improve the rate of early screening of heart diseases.This paper designs and implements a set of heart sound acquisition,transmission,storage and display and heart sound signal analysis system,including hardware and software part.System hardware: the original heart sounds signal the large amplitude at about 40 m V,which is based on medical auscultation head and electret microphone as a heart sounds sensor can be true to extract the complete heart sound signal,and then the signal conditioning circuit will be weak signal amplification,filtering processing,STM32F103ZET6 is used as the main control chip,and the 12-bit ADC controller is used to carry out AD sampling of 2500 Hz heart sound signal,wireless transmission adopts Bluetooth 5.0 module with ultra-low power consumption and low delay n RF52832 as the core,which receives data from UART and then wirelessly transmits it to PC.The electronic auscultation interface is designed on the PC based on MATLAB platform,which can realize the real-time display of heart sound waveform,data storage and playback.Heart sounds denoising part,in view of the heart sound signals are weak,strong background noise of non-stationary random low frequency signal,IIR low-pass digital filter can be a good filter to remove high frequency noise,IIR band-stop filter to remove 50 Hz frequency interference,compared with the experiment in the denoising effect at the same time,using chebyshev type?low-pass filter order is far less than butterworth low-pass filter.In view of the noise and signal overlapping spectrum,based on the time-frequency locality and multi-resolution characteristics of wavelet denoising,db6 wavelet global default threshold denoising is used to denoise the signal in five layers db6 wavelet global default threshold denoising is used to denoise the signal in 5 layers.Normalized Shannon energy method and Hilbert-Huang Transform method were compared to extract the envelope of heart sound.After preprocessing,db6 wavelet basis is selected to decompose heart sound signal into wavelet packet tree.The experiment found that when 32 frequency bands were obtained by wavelet packet 5-layer Shannon entropy decomposition,by analyzing the energy spectra of normal and abnormal heart sound signals,the node energy values of the first 16 frequency bands of the decomposed signals can reflect the changes of the components of normal and abnormal heart sound,and E1?E16 can better reflect the difference between normal and abnormal heart sound signals.The fully connected neural network was selected as the classifier,the 16 frequency band energy eigenvalues obtained by wavelet packet decomposition were used as the input of the classifier.A total of 470 samples of heart sound data from Physio Net database and normal human heart sound data collected by our own research equipment were used to train and test the fully connected neural network and optimize the parameters of the classifier.During the training,choose the 430 samples were crossverified by 6 folds,and the average accuracy of the 6 groups of models was 89.53%.During the test,the rest of 40 normal and abnormal heart sound samples were selected and put into the 6 groups network to test respectively,the average recognition rate was86.7%.The heart sound research system in this paper is helpful to the establishment of the heart sound database,and the FCNN algorithm adopted in the recognition of normal and abnormal heart sounds has also achieved good effects,which is expected to be applied in the auxiliary diagnosis of heart sounds.