Neural Network ECG Portable Monitoring System Based On High Permeability Electrode

With the rapid development of modern society,the aging of the population and the change of life style make cardiovascular disease become the main factor of human health.The medical concept of cardiovascular disease based on prevention has become one of the main development modes to guarantee human health.Therefore,it is of great significance for the early prevention of cardiovascular diseases to actively develop portable,real-time diagnosis technology of ECG（electrocardiogram）signals.Hence,for the mass prevention and health field,our project is committed to the realization of portable and real-time monitoring of ECG signals.However,in terms of portable and real-time monitoring,there are many problems in the present characterization and diagnosis system of ECG signals.From the perspective of engineering,there are three main problems:1.Poor air permeability of the traditional Ag/Ag Cl electrode leads to the accumulation of human sweat,which affects the stability of ECG signal characterization;2.low integration of connections between circuit and electrodes;3.low portability and real-time diagnosis degree of traditional ECG diagnosis system.According to the above problems,the main contents of this paper are summarized as follows:Firstly,this thesis designs the ECG signal characterization system,including designing of high permeability ECG electrodes,designing of integrated ECG sensor circuit and its interface.In the design of breathability of ECG electrodes,this study uses porous TPU（Thermoplastic polyurethanes）films as the substrate,and Ag nanosheets-carbon nanotubes composite material as the induction electrode.By hot pressing curing,this project achieves electrodes with high air permeability of 23mg·cm^-2·h^-1,high signal-to-noise ratio of 10 d B,as well as the similar impedance with traditional electrodes.As for ECG signal processing circuit,this paper makes the flexible circuit based on PI（Polyimide）,and seals the circuit with Eco Flex Gel.Finally,through finite element analysis and physical test,this thesis illustrates the mechanical robustness and reliability of the ECG signal characterization system in terms of the portable application.Secondly,this thesis deals with the ECG diagnosis system including high accuracy neural network designing,and low delay Field Programmer Gate Array（FPGA）implementation.In the design of neural network,this thesis proposes an end-to-end ECG diagnosis model architecture under inter-patient mode using the MIT-BIH database.In the test of this ECG classification model,the diagnosis accuracy of ECG signals reached 89.93%,among which the F1 value of class N and class V reach 94.19 and 86.46 respectively.In the FPGA deployment,a Zynq-XC7Z010 logic system（XILINX）is deployed the neural network because of practicability.And then,this thesis achieves the ECG diagnosis system with small model space（～810K）,short classification time（～0.38 ms）for portable applications.To sum up,this thesis carries out around portable and real-time monitoring to design a high air permeability of 23 mg·cm^-2·h^-1,high signal-to-noise ratio of 10 d B electrode for ECG representation system.According to the ECG diagnosis system,this thesis achieves low delay of 0.38 ms and high accuracy of 89.93%.In conclusion,the work has great potential in the field of wearables devices and mass prevention.