The Research Of Fetal Ecg Signal Extraction Method And Application

Fetal monitoring is very important in perinatal period, and fetal ECG is one ofthe most effective methods of fetal monitoring. As fetal ECG reflects the completeview of fetal heart activities, fetal acidosis and anoxia can be discovered by theanalysis of fetal ECG waveform. The problem is that fetal ECG in maternalabdominal signal acquired by non-invasive electrode method is low in SNR, and isusually contaminated by strong maternal ECG. So extracting fetal ECG from thematernal abdominal signals is the focus of fetal monitoring. Many fetal ECGextracting methods have been reported, such as, coherence average, adaptive filtering,independent component analysis and neural networks, etc. However, these existingmethods have various disadvantages, such as, low accuracy or complex calculation. Inthis paper, the algorithm of fetal ECG extraction was analysed in detail, and two newfetal ECG extraction methods were proposed, the primary contents are as follows:(1) Wavelet analysis and adaptive filtering are both commonly used methods forsignal analysis. These two algorithms were combined in this paper for fetal ECGextraction. After wavelet decomposition of the signals both from maternal abdomenand chest, LMS algorithm is introduced at each level of the wavelet coefficients, andfetal ECG can be obtained by wavelet reconstruction. Fetal ECG extracted by thismethod is clear and neat. Maternal ECG and noise disturbance are eliminated bettercompared with LMS algorithm.(2) Considering the phenomenon that maternal ECG and fetal ECG havedifferent characteristics in the wavelet transform, wavelet threshold algorithm waspresented in this paper to extract fetal ECG. After wavelet decomposition of maternalabdominal signal, proper threshold was used in the wavelet coefficients to eliminatecomponents of fetal ECG. Then maternal ECG and fetal ECG in maternal abdominalsignal can be distinguished. The proposed algorithm was tested by various simulatingdata and246clinical data from3different databases, and a test-report was finishedafter result analysis. This method based on single-channel signal is not affected bycorrelation of multi-channel signals, and it also reduced computation compared withwavelet transform modulus maximum for fetal ECG extraction.(3) Real-time of the wavelet threshold algorithm used for fetal ECG extractionwas analysed. The time complexity of this algorithm was described by formula T(n)=O(N2), and the calculation would be complex as length N of the signal increase.Suggestion of replacing wavelet decomposition and reconstruction by FFT waspresent to reduce computational complexity and improve real-time performance, bywhich the time complexity could be reduced by O(N2) to O(Nlog2N).