ECG Signal Recovery And Real-time Diagnosis In Wireless Body Area Networks Based On Compressive Sensing

With the rapid development of modern society,people pay more and more attention to health in life.Correspondingly,Wireless Body Area Networks has attracted more and more people's attention.People began to develop related technologies to monitor the physiological parameters of patients.In the face of cardiovascular diseases,its outbreak requires that monitoring system should be real-time and stable,and usually requires data compression for higher energy efficiency.In recent years,a large number of studies have verified that Compressive Sensing is an effective solution.The integration of this theory with the Wireless Body Area Networks makes telemedicine more intelligent.The traditional compressive sensing method works well under the condition of steady data.However,the effect is greatly reduced in the face of real-time physiological parameters.Therefore,this thesis studies the ECG signal reconstruction and real-time diagnosis of Wireless Body Area Networks based on compressive sensing.On the theoretical framework of compressive sensing,adaptive control strategy is designed in wireless body-area network transmission system.At the same time,the reconstruction algorithm is optimized according to the characteristics of ECG signal and the real-time transmission,so as to achieve double optimization of reconstruction accuracy and compression performance.In order to transmit the data more flexibly,we need to set up the mathematical model for the sensing node,the coordination node and the wireless transmission channel in the Wireless Body Area Networks,and design the data compression rate by using the feedback control thought.We add the correlated sparsity compression rate estimation module in the sensing node,which is designed to meet the real-time conditions,as much as possible to reduce the compression rate,improve the ECG signal reconstruction accuracy.In the correlated sparsity compression rate estimation module,the frequency domain analysis method is used to reveal the dynamic characteristics of the ECG data and to model the relationship between the compression rate and the sparsity.In addition,the fast K-Means clustering algorithm is used to classify the ECG signals into two groups according to the sparsity.The models are simplified,fitted with piecewise nonlinear functions,and the problem is transformed into a linear programming problem.In the packet loss compensation module,the closed-loop module and control process are designed,the compression strategy is formulated.The compressive sensing parameters are updated and estimated based on the channel state and data sparsity,in order to reduce the impact of packet loss and lay the foundation for real-time high-precision feature recognition and remote diagnosis of ECG signal.At the same time,the reconstructed algorithm of the coordination node is improved,and the Subspace Pursuit reconstruction algorithm is selected and optimized to make it become the high precision and fast,named GBSP algorithm.Under the conditions of lower data compression rate,the improved GBSP algorithm can still obtain the reconstructed signal with medical value,that is,the reconstruction error PRD is below 9%.Based on the idea of Gradient Boosting in machine learning,we search for the parameter K in the algorithm and search for the direction of the gradient descent to re-establish the model for the residual.In the condition that the residuals can not be neglected,more than K sparsity atoms are selected to enter the candidate set and then backtracked.After two turns of optimization,the GBSP algorithm can control the reconstruction error PRD under 9% at most with a lower compression ratio such as 30%,and the reconstruction capability is very satisfactory.To meet the needs of real-time monitoring in telemedicine systems,we have designed a novel online ECG signal recognition algorithm to locate QRS complex with high precision and obtain Q wave,R wave and S wave spot.We make ECG signal through the window integrator,and the purpose of calculating the rolling average is to further smooth the ECG time-domain signal;and then for each data point using the related threshold to determine the current flag state,we identify the waveform which meets a certain amplitude and time threshold range.In order to improve the robustness of the algorithm,an adaptive mode is applied to the amplitude threshold corresponding to the detected waveform.Also,we use two sources of ECG data,including the MIT-BIH database and actual hardware acquisition to test and compare the accuracy of the two algorithms to identify ECG signal.Further,using the designed wireless network transmission system,the reconstructed ECG signals are respectively identified with or without packet loss,which further verifies the feasibility and robustness of the designed system.