The Research On Noninvasive Fetal Electrocardiogram Signal Extraction Methods

Fetal ECG which reflects the fetal heart activity has an irreplaceable role in fetal health monitoring and disease diagnosis.As the fetus ECG signal from the abdomen of the maternal body is very weak and is easily disturbed by various noises,it becomes a hot issue how to extract the pure fetal ECG signal.Fast independent component analysis(FASTICA)algorithm has become a hot research topic in the field of fetal ECG extraction due to its simplicity,small amount of calculation and wide application range.The traditional FASTICA initial random selection method will cause the algorithm to fall into the local optimal value,the extracted fetus ECG signal often mixed with maternal ECG signal.In addition,its iterative efficiency of fetus ECG signal extraction is low.For the disadvantages mentioned above,we proposed a Terrific-FASTICA method to extract FECG,e.g.the steepest descent method to set the initial value of FASTICA,and replace the traditional iteration method of FASTICA by chordtruncation method.We compared the T-FASTICA and FASTICA by artificial synthetic signal and international databases signal.In artificial synthetic signal separation experiments,results showed verify the T-FASTICA extraction of independent components,and its extraction better,faster extraction.In semiartificial ecg signal experiment the extraction results of signal to noise ratio,the source signal similarity,computing efficiency and other performance indicators verify the T-FASTICA fetus ECG signal extraction feasibility,In fetus ECG signal extraction experiments,two international databases(Dalay database and PCDB database)were carried out to verify the stability and reliability of TFASTICA in fetus ECG signal extraction.Finally,fetus heart rate testing was performed on extracted fetal ECG signals.Experiments demonstrated that the new T-FASTICA-based approach outperformed QIO-based methods(the best known method of fetal heart rate detection)in fetal heart rate testing.