A U-Net-Based Deep Learning Method for R-Wave Detection in Wearable Device

In the presence of noise or abnormal cardiac waveforms, achieving accurate QRS localization and heart rate estimation remains challenging, particularly when the ECG signals are acquired by wearable dynamic ECG monitoring devices. To address this issue, we propose a deep learning model based on the U-Net architecture. This method aims to improve the accuracy of R-wave detection, especially under conditions of noise or waveform abnormalities. The model was trained and validated using the the China Physiological Signal Challenge (CPSC) 2019 database, and its performance was further evaluated on the MIT-BIH Arrhythmia Database. Experimental results show that the proposed algorithm achieved a heart rate calculation accuracy of 0.9421 on the CPSC 2019 non-public test set, and reached a sensitivity of 99.88% and a positive predictive value of 99.90% on the MIT-BIH database. The cross-database tests showed high accuracy and robustness, making it suitable for wearable device applications.