Efficient ECG Arrhythmia Detection on FPGA using Machine Learning and Fiducial Windowing

This work presents an efficient FPGA-based system for real-time detection of ECG arrhythmias using machine learning and fiducial windowing techniques. The proposed system integrates FPGA hardware acceleration to achieve low latency and high energy efficiency while maintaining superior classification accuracy, making it well-suited for portable health monitoring devices. ECG signals are preprocessed with a Butterworth filter to remove noise, followed by feature extraction through Discrete Wavelet Transform (DWT). The fiducial windowing method identifies key ECG components such as the P-wave, the QRS complex, and the T-wave, allowing the extraction of clinically relevant features. These features are then classified using a machine learning model implemented on an FPGA, allowing for rapid and accurate arrhythmia detection. The hardware-based solution significantly outperforms traditional software implementations in terms of real-time performance and power consumption. The proposed system achieved an impressive accuracy of 99.7%, a processing speed of 0.723 s, and a power consumption of 0.42 mW. The design was implemented using Xilinx Vivado 2022 EDA tools on the Xilinx PYNQ FPGA platform. This study demonstrates the potential of FPGA-based machine learning systems for efficient and reliable real-time ECG analysis, paving the way for advanced wearable health monitoring applications.