Research On ECG Reconstruction Algorithm Based On Sparse Bayesian Learning

Recently,the incidence of heart disease in China has been increasing year by year due to the increasing age of society.It has become a growing concern to detect abnormal signs in the human body in time and avoid sudden heart attacks through medical interventions.Body Area Network technology provides a solution for remote real-time monitoring of ECG signals,blood pressure data and other indicators of cardiac patients.Due to limitations of device power consumption and information transmission,the information should be compressed on the sampling side and recovered via reconfiguration algorithms on the receiver side.With the diversification and widespread use of human information acquisition devices,the amount of collected information is growing rapidly which brings higher requirement to the information reconstruction efficiency.Therefore,this paper investigates the reconstruction algorithm of ECG signal and focuses on efficiency of the algorithm.In order to achieve improvement of reconstruction efficiency of ECG signals on the receiving side,an Expectation Propagation-Sparse Bayesian Learning algorithm is introduced in this paper.On the one hand,the Expectation Propagation algorithm has the advantage of high operational efficiency and high accuracy of signal reconstruction,which can be used to improve the efficiency of signal reconstruction.On the other hand,the Expectation Propagation algorithm should preserve the critical information of the signal,which is actually impossible in real scenarios since the priori distribution of ECG signal is unknown.So the Expectation Propagation algorithm cannot be used alone.This paper introduces the Sparse Bayesian Learning algorithm to the process of the priori distribution of ECG signal,extending the usage of the Expectation Propagation algorithm to solve the generalized linear problem,and eventually achieving the reconstruction of ECG signal by the Expectation PropagationSparse Bayesian Learning algorithm.To measure the performance of the algorithm,this paper adopts four indicators:computing time,PRD,MSE(mean square error),SNR(signal-to-noise ratio).This paper conducts several simulations against not only this algorithm,but also against some mainstream reconstruction algorithms such as Orthogonal Matching Pursuit algorithm,BSBL-EM and BSBL-BO.The simulation results show that the Expectation Propagation-Sparse Bayesian Learning algorithm has better performance of efficiency over algorithms such as Orthogonal Matching Pursuit algorithm and block sparse Bayesian Learning.In addition,with signal compression ratio of 0.3,0.5,and 0.7,the Expectation Propagation-Sparse Bayesian Learningalgorithm shows good performance in terms of PRD,MSE,SNR.The reconstructed signals can accurately reflect the critical features of ECG,thus meeting the requirement of medical diagnosis.In summary,Expectation Propagation-Sparse Bayesian Learning algorithm is capable of achieving reconstruction efficiency on ECG signals with high reconstruction accuracy.With the increasing growth of wearable devices,the Expectation Propagation-Sparse Bayesian Learning algorithm can greatly improve the reconstruction efficiency at the receiver side,reducing the risk of data loss and improving the speed of response to abnormal situations,which is of vital importance to our society.