Fetal ECG Extraction Based On RobustICA And Its Improved Metrod

FECG has a wide range of clinical applications as effective indicators for fetal health.Non-invasive method is the main way to get FECG, which brings maternal ECG and noise at thesame time, so to extract FECG from the mixed-signal is the premise of the medical testing.The paper carries on the following aspects:Firstly, the paper applies RobustICA to the extraction of FECG for the first time. Based on thestudy of RobustICA algorithm and comparation with FastICA through the relevant test, the speedand accuracy of separation of RobustICA is confirmed to be superior to FastICA, and the algorithmshows great robustness to data deficiencies.Secondly, combining with wavelet denoising, the thesis firstly improves RobustICA after acareful study of maternal and fetal ECG characteristics. Maternal-fetal ECG signals are firstlypreprocessed with wavelet denoising, and then separated by RobustICA. The clarity of extraction isgreatly improved as results show.Thirdly, the thesis improves RobustICA with wavelet decomposition from a new perspective.After simplifying the mixing ECG signals with wavelet decomposition, the complex signals areconverted to simple wavelet coefficients, which enhancing the overall separation effect of thealgorithm.Fourthly, the paper makes the first attempt to improve RobustICA with sparse decomposition.Under the use of matching pursuit optimized by genetic algorithm, the paper sparses the mixingECG, and then separates the singals with RobustICA. The algorithm improves the separation speed,and ensures the separation accuracy at the same time.