| The fetal electrocardiogram(FECG) reflects the electrical activity of the fetal heart, including the heart rate, wave forms and some other dynamic information of the fetus. The analysis of FECG may be helpful to understand the fetal health state and monitor the present heart disease. Abdominal FECG is detected by Non-invasive measurement is not harmful but convenient. The abdominal FECG is always mixed with maternal electrocardiogram(MECG) and noises like Baseline Wander(BW) which make it uneasily to observe or analysis the FECG. The extraction of clear FECG from the mixed FECG obtained from abdomen is a difficult and hot issue of ECG monitoring, which is under researching in this object.In this paper, ECG model is proposed firstly, then mixed ECG model is presented to synthesize the mixed ECG including MECG, FECG and noises in it. The mixed ECG model is helpful to understand the morphology and verify the performance of the algorithms of the FECG extraction. Thinking about that FECG gathered from maternal abdominal wall contains MECG and noises, the dynamic ECG model need to approach the real mixed ECG in the frequency, amplitude and morphology domain.To extract clear FECG from mixed ECG signals got from maternal abdominal wall, in this dissertation, three algorithms based on Blind Source Separation(BSS) are adopted: Joint Approximate Diagonalization of Eigen-matrices method(JADE), Fast Independent Component Analysis method(Fast ICA) and Periodic Component Analysis based on generalized eigenvalue decomposition method( ?CA).To verify the validity of ECG extraction, data from DaISy and data synthetized by ECG model are chosen to be used for analysis of signal extraction. The results show that all the three algorithms can extract FECG. For further analysis between three algorithms, algorithms are evaluated from aspects of visual observation, signal to noise ratio. The analysis shows that ?CA has a higher SNR and the FECG extraction goes better obviously, while the FastICA algorithm performs a little bit better than JADE algorithm. |