A Research Of ECG Collection And Arrhythmia Detection System Based On CAPSNET

Heart disease is a serious threat to people’s health.Electrocardiogram can truly reflect the health condition of the heart.It is one of the commonly used method to clinically detect heart disease.The traditional electrocardiograph machine can effectively detect heart disease,but it cannot meet the needs of daily life detection due to it is lack of portability.The existing methods of ECG automatic detection are limited by feature extraction,and the recognition accuracy still need to be improved.Aimed at the improvement of convenient collection and automatic detection,this paper proposes a Wearable ECG acquisition and arrhythmia detection system based on CAPSNET.The research work in this paper is divided into two parts:A wearable electric acquisition device is designed.Based on STM32 and AD8232,the hardware circuit is built to collect ECG signal and the system is controlled through software programming.According to the spectral characteristics of the ECG signal,the collected ECG signal adopts two digital filtering methods to suppress the noise.The adaptive filtering algorithm is used to denoise the electromyographic noise in the ECG signal while the wavelet transform algorithm to suppress the baseline drift.The processed ECG signal is transmitted to the host computer via Bluetooth.The CAPSNET model is constructed to detect cardiac arrhythmias.The CAPSNET model consists of eight layers.Through the feature learning and mapping by layer by layer,the model obtains the deep features of ECG signal,and finally cardiac arrhythmias classification is achieved.The method which encapsulating features into vectors improves the robustness of the model,completes the classification of arrhythmias.Based on the Tensor Flow framework,a CAPSNET model for detecting arrhythmia is built.The data of training samples are established by using the data from the MIT/BIH arrhythmia database,and the model is trained with these data.Through the fine tuning of dynamic routing algorithms,a global optimal model is obtained.In the experiment,the ECG signal is collect by the designed device,and the test data from the MIT/BIH arrhythmia database is extracted to evaluate the CAPSNET model,and the classification accuracy of five heart rhythms are obtained.The results show that the system can meet the requirement of convenient collection,and the CAPSNET model of the system has good classification performance for arrhythmia.