Design And Implementation Of An Electrocardiographic Signal Analysis System Based On Neural Network

With the development of the economy and the changes in people’s lifestyles,the number of cardiovascular diseases in China continues to increase.Cardiovascular and cerebrovascular diseases have become the first cause of death.The prevention and treatment of cardiovascular diseases is urgent.Electrocardiogram(ECG)is used to characterize the electrical activity of the heart and is an important basis for doctors to diagnose cardiovascular diseases.This paper studies an ECG signal analysis system based on deep learning convolutional neural network.The main work of this paper is divided into three parts:1.Propose an ECG signal QRS complex detection algorithm based on energy segmentation.Use MIT-BIH arrhythmia database(MITDB),European ST-T database(ESTD),motion artifact contaminated ECG database(MACECGDB)and portable ECG signal detection equipment to conduct simulation experiments to verify the algorithm Effectiveness.Experimental results show that the overall accuracy of QRS feature detection is 99.14%,the sensitivity is 99.36%,and the positive detection rate is 99.78%.The QRS complex detection algorithm of ECG signal based on energy segmentation has strong real-time performance and good noise robustness.It can also detect the QRS position more accurately for ECG signals containing motion artifacts,and its performance is better than other algorithms.2.Propose a mixed model structure of arrhythmia with convolutional neural network(CNN)as the core.On the basis of the traditional CNN structure,the spatial pyramid pooling layer(SPP)and the extreme learning machine(ELM)are introduced to improve the classification accuracy and reduce the training time.The SPP 3-layer pyramid structure can process variable-length ECG signal fragments on the one hand,and on the other hand,different pooling steps can perform secondary mining on the features extracted by the convolutional layer;ELM can effectively replace the original Softmax classifier Shorten training time.A simulation experiment was performed using the MIT-BIH arrhythmia database to verify the accuracy of the mixed model 4classification.The experimental results show that the overall classification accuracy on the test set is 99.16%,the sensitivity is 99.85%,the specificity is 98.89%,and the precision is 99.85%.3.Realize the design of arrhythmia classification system.A portable ECG signal detection device is designed with the BMD101 chip of Neuro Sky as the core for daily detection.A graphical user interface is designed by using MATLAB GUI,which can save the ECG signal obtained by ECG signal detection equipment to the specified folder.It only needs to click the interface button to realize the display of heart rate calculation and arrhythmia classification.This ECG signal analysis system is small and easy to operate.