A novel ECG segmentation for compression using Fourier series approximation in e-health devices

摘要

Since ECG signals capture the conduction of the heart, physicians monitor their patients' using special equipments. Because the monitoring lasts for long-time periods, the device should have a reasonable lifetime. Therefore, the recorded signal is manipulated in compressed format, while preserving the diagnostic information of the signal. In this paper, a novel segmentation of electrocardiogram signals is proposed for compression by Fourier series. A set of significant turning points is computed to strip the signal into sharp-peaks segments that limit the use of Fourier approximation. Two datasets are used for testing the algorithm. The percentage root-mean difference (PRD) measure and the weighted diagnostic distortion (WDD) are used to report the results. The method has superb performance at all bit-rates, and good quality score (QS). With some constraints on device's architecture, the algorithm can be implemented and achieves a high compression ratio while preserving the diagnostic features of the signal.