Research On ECG Compressed Sensing Of Integration Oriented

Compressed sensing has been deeply studied and developed in the field of ECG remote health monitoring.Traditional methods are based on a priori measurement matrix and iterative optimization to achieve compressed sensing of ECG signals.However,the iterative optimization process will cause low performance,and the prior knowledge of measurement matrix limits its application in sundry compression ratio scenarios.Therefore,it is the key to propose an efficient method applied to variable compression ratio scenarios.In the dissertation,a method of multi rate ECG compressed sensing based on convolution integrated is proposed,it is described by theoretical model and algorithm implementation.Based on compressed sensing theory and neural network,the theoretical model constructs a compressed cooperative network framework composed of compression and reconstruction,and proposes a multi rate implementation of compressed network in the form of convolution integration.The compressed network architecture is established in block way,and the reconstruction network architecture is introduced by stacking hourglass residuals.The algorithm implementation puts forward the compound loss suitable for compressed cooperative network framework,expounds its parameter configuration and training method,and analyzes its design parameters.In this dissertation,a number of test experiments have been carried out on MIT-BIH arrhythmia dataset,including comparative test and characteristic test.Compared with other methods,method in this dissertation has higher reconstruction accuracy and efficiency.The characteristic test analyzes the advantages of convolution integration multi rate implementation,which requires less training time and higher reconstruction accuracy compared with multiple single rate combinations.