RA-LENet:R-Wave Attention and Local Enhancement for Noise Reduction in ECG Signals

Cardiovascular disease is a major life-threatening condition commonly monitored through electrocardiogram (ECG) signals. However, the ECG signals currently generated by sensors are often accompanied by a plethora of diverse types of noise with different intensities, which causes a lot of interference in downstream tasks. In this work, we propose a deep learning based method for ECG signal denoising. Due to the different frequency characteristics of different types of noises, we use a Transformer with local enhancement as a feature extractor which can capture global dependencies. In addition, we introduce an R-wave attention mechanism to improve the most difficult R-wave reconstruction. Our experimental results demonstrate the effectiveness of our approach in denoising different types of strong noises, outperforming the state-of-the-art (SOTA) methods. The code is available at https://github.com/caprilovel/ECG_Denoise.