Extract fetal ECG from single-lead abdominal ECG by de-shape short time Fourier transform and nonlocal median

摘要

The multiple fundamental frequency detection problem and the sourceseparation problem from a single-channel signal containing multiple oscillatorycomponents and a nonstationary noise are both challenging tasks. To extract thefetal electrocardiogram (ECG) from a single-lead maternal abdominal ECG, weface both challenges. In this paper, we propose a novel method to extract thefetal ECG signal from the single channel maternal abdominal ECG signal, withoutany additional measurement. The algorithm is composed of three mainingredients. First, the maternal and fetal heart rates are estimated by thede-shape short time Fourier transform, which is a recently proposed nonlineartime-frequency analysis technique; second, the beat tracking technique isapplied to accurately obtain the maternal and fetal R peaks; third, thematernal and fetal ECG waveforms are established by the nonlocal median. Thealgorithm is evaluated on a simulated fetal ECG signal database ({\em fecgsyn}database), and tested on two real databases with the annotation provided byexperts ({\em adfecgdb} database and {\em CinC2013} database). In general, thealgorithm could be applied to solve other detection and source separationproblems, and reconstruct the time-varying wave-shape function of eachoscillatory component.