| Fetal electrocardiogram(FECG) contains much important information about the health of the fetus,which is the most recurrent source of signal of fetal cardiac activity. Detection of fetal ECG therefore has become significant in the clinical application of fetal monitoring. The FECG signal can be collected from the abdomen of pregnant women or electrodes placed on the fetal scalp. The former has advantages over the latter in convenience of operation, non-invasion, monitoring persistence. But the FECG signal collected from the abdomen of pregnant women is very weak, of low SNR and interfered by the maternal signal. So the extraction of pure FECG is an important research topic in biomedical signal processing.The feasibility of the blind source separation(BSS) algorithm applied to the extraction of FECG is demonstrated,which is combined with the characteristics of the fetal ECG itself. And the model of second-order statistics of blind source separation is applied to extract the FECG. The specific content of the work is following.Blind source separation based on second-order statistics makes the second-order statistics of the sample data and structural characteristics of the source signal timing as the basis to achieve blind source separation. The R wave of FECG provides an important basis to determine the timing of the structure. But the R wave of FECG is difficult to detect because of its characteristics. The difference between maternal modulus maxima and fetal modulus maternal emerged after wavelet transform. In this paper, an algorithm based on the binary wavelet transform and modulus maxima are provided for detection R wave of FECG by classifying the sequence of modulus maxima of the fetal and the maternal in form of modulus maxima blocks. After submitting this method to the MIT Physio Net/Cin C2013 challenge contest, the evaluation from the results feedback shows R-wave accuracy rate reached 93.8%.The periodic structure of FECG can be described accurately by the R wave of FECG. The algorithm periodic component analysis(Ï€CA) based on the periodic structure of FECG is excellent to extract the FECG. But the extraction of FECG based on periodic component analysis can be partial failed because of the similarity of periodic structure between the maternal ECG and the FECG. In view of this, the best of the time-delay estimation of the covariance matrix improved. Then the feature vector, which is corresponding to the maximum eigenvalue resulting from the generalized eigenvalue decomposition to the matrix pair of autocorrelation matrix and delay correlation matrix, is estimated FECG. And the algorithm which is improved and AMUSE, SOBI, periodic component analysis are applied to extraction FECG from the given data set which is from the MIT Physio Net/Cin C2013 challenge. The result shows the effect of extraction of FECG based on our algorithm is remarkable. |
