A Wearable Flexible ECG Monitoring System Based On Mobile Terminal Analysis

As the pace of society accelerates,public health is increasingly threatened,especially in the field of cardiovascular.However,continuous electrocardiograph(ECG)monitoring has encountered a significant bottleneck since the device is not portable and lack real-time analysis capability.Therefore,a smart and wearable ECG monitoring system is required.In this thesis,a wearable flexible monitoring system was implemented,which has two main components.Generally,a flexible interconnection circuit was innovatively proposed to collect ECG and the neural network on the mobile was utilized to analyze in real-time.In hardware,according to the characteristics of the ECG,the chip AD8233 was selected and its peripheral circuits were designed for signal amplification and filtering.Then,the integrated chip CC2541 was selected to meet the requirement of wireless communication.Its communication digital circuit and programming were implemented to realize analogue-to-digital conversion and Bluetooth transmission.To solve the wearability issue,the volume of each circuit was reduced and a new type of screen-printed flexible wire was designed to combine traditional silicon-based circuits with flexible circuits.As a result,the flexibility of the entire circuit was maintained without sacrificing signal quality,which achieves the goal of stable and wearable ECG measurement.In terms of software,a neural network model suitable for mobile terminals was trained and a dedicated Android APP was developed aiming to obtain device intelligence with portability.Then the miniaturized ECG analysis neural network model was embedded into the mobile application based on the mobile terminal framework TFLite and model Mobile Net,which can realize human-computer interaction,wireless communication,dynamic ECG display and real-time classification.The total area of circuit boards is only 9.8cm~2 and the SNR of ECG reaches 40d B.After flexible interconnection,the integrated circuit can be bent nearly 180°.The"inter-patient"classification model is adopted for intelligent analysis based on ECG segments.The trained neural network model has 22 layers but only 2.4 million parameters,whose average accuracy is 87%and the classification accuracy of N-type and V-type ECG is almost 92%and 91%.After converted to tflite format,the model is only 9 M and takes less than 30 ms to run an intelligent ECG diagnosis on the mobile.The wearable system achieves real-time ECG measurement and portable intelligent diagnosis,which is significant for the subsequent development of wearable ECG monitoring device.