An improved adaptive periodical segment matrix algorithm for ECG denoising based on singular value decomposition.

Abstract

Wearable devices that monitor heart health of cardiac disease patients in real time are in great demand. We propose an algorithm of improved segment periodical matrix construction for irregular electrocardiogram (ECG) signal denoising. While splitting the heartbeat based on each RR interval for periodical segments matrix construction, the as-filtered ECG signal is reconstructed by the maximum singular value after a singular value decomposition. The results demonstrate a higher noise reduction effect with lower signal distortions of our methods compared to several singular value decomposition counterpart approaches. Our method has great potential to enhance wearable devices diagnosis accuracy by denoising the complex noises such as electromyography artifacts in real-time ECG sensing.