Intelligent ECG Monitoring System Based on DenseNet with Portable Deployment

To address the challenges of real-time analysis and portability in cardiovascular disease diagnosis, this paper proposes an intelligent ECG monitoring system based on densely connected convolutional neural networks. First, the MIT-BIH database is used to divide the ECG signals into five categories: N, V, S, F, and Q according to the AAMI standard, and the SMOTE method is applied to address the data imbalance problem. Secondly, a DenseNet network model is constructed to achieve end-to-end classification of 5-second ECG signal fragments. Finally, the trained model is deployed onto a Raspberry Pi 4B hardware platform. Experiments show that the system achieves a global classification accuracy of 90.38%. The precision rates for N and V categories are 95.15% and 88.35%, respectively. Each signal is processed in 0.3 seconds, which satisfies the requirements for real-time clinical monitoring.