Research Of Heart Sound Reconstruction Based On CS And BSBL

With the development of wireless sensor technology,body domain network as a new type of cross technology arises at the historic moment.It realizes real-time monitoring to body health through the body of the sensor nodes and wireless transmission technology.In the process of monitoring data,large-scale physical datum must be compressed reconstructed,which leads to large amount of data,the transmission rate slow,low precision in the terminal.The heart sound is one of the important physiological signals,it can reflect the cardiovascular disease.Therefore,the topic is research of Heart Sound Signal Reconstruction Based on Compressed Sensing and Block Sparse Bayesian Learning.The thesis mainly body area network as background,combined with compressed sensing,implement under-sampling of heart sound signal,the measurements for data transmission,on the remote monitoring decoding reconstruction signals.The thesis explores that the traditional compression sensing reconstruction algorithm and block sparse Bayesian learning algorithms apply to reconstruction of heart sounds.To design interface will result heart sound data visualization by MATLAB GUI;and to evaluate the performance algorithms and heart sounds normal or abnormal;ultimately achieve the heart sound datum processing on body area network.In order to complete the above objectives,the thesis mainly studied in-depth from the following aspects:(1)Heart sound body area networkTo establish compressed sensing reconstruction system of heart sound based on body area network.In the heart sound signal body area network,combined with block sparse Bayesian learning,heart sound signal can be sparse,the measurements got by sub-Nyquist sampling and be transmitted to the remote monitoring decoding end.Finally,heart sounds were reconstructed via algorithms and visualization display.(2)Compressed sensing analysisCompressed sensing is a new theoretical framework for information acquisition and processing,mainly studied in three aspects: the sparse representation of signals,measurement matrix structure,reconstruction algorithm.The heart sound signals have sparse by discrete cosine transform(DCT),measurement matrix compressed signals,selected orthogonal matching pursuit algorithm(OMP),which reconstructed clinical normal and abnormal heart sounds,and analysis the results of reconstruction by indexes running time,signal to noise ratio,percentage root-mean-square difference,structural similarity.(3)Reconstruction of heart sounds by block sparse Bayesian learningIn order to solve low heart sound signal reconstruction accuracy,the algorithm running time of slow issue;the thesis combined block sparse Bayesian learning with compressedsensing in clinical heart sound signals.the block heart sound signals were processed and compressed via a sparse binary matrix.In the process of reconstruction,heart sounds were reconstructed by block sparse Bayesian learning-bound optimization and block sparse Bayesian learning Expectation Maximization.The thesis compared orthogonal matching pursuit algorithm with block sparse Bayesian learning-bound optimization and block sparse Bayesian learning Expectation Maximization by indexes.The experimental results showed that the proposed compressed sensing method which can acquire high reconstruction accuracy,process large amount of heart sound data,and run fast.(4)System interface design,data visualizationThe system makes heart sound visualization by MATLAB GUI.The thesis design a system for heart sound reconstruction based on compressed sensing and analysis heart sounds.The system evaluated OMP,BSBL-EM,BSBL-BO algorithms reconstruction performance and efficiency of the algorithm via the compression ratio,PRD,SSIM,running time;from the heart rate,TD/TS,S1/S2 physiological indicators to assess cardiac reserve capacity and judge heart sounds normal or abnormal.The GUI makes visualization of heart sound waveform and shows the reconstruction algorithm performance,user-friendly operation.Although Block sparse Bayesian learning methods for heart sound signal in the reconstruction accuracy of heart sound signals,large amount of data,the algorithm efficiency achieved satisfactory results.This article will focus on the improvement of future clinical heart sounds dynamic database and data transmission for further in telemedicine,the analysis of heart sounds pathology and the transmission of heart sound data work on body sensor network.