Study On Extraction Of Fetal ECG Signal Based On LSSVM

Fetal ECG monitoring is an important way to monitor fetal health during pregnancy.It can effectively detect fetal growth and development in the mother's body,and responds in a timely manner,which can reduce the morbidity and mortality of the fetus.Therefore,it is of great theoretical significance and practical value to effectively extract clear fetal ECG signals.Due to the complexity of the maternal fetal growth environment,fetal ECG acquisition has a weak,non-stationary nature,and is often affected by maternal ECG,power frequency interference,baseline drift and myoelectric interference.The fetal ECG with less energy is not easy to be observed in the original ECG signal,which makes it more difficult to extract fetal ECG features and limits its clinical application.At present,using wavelet decomposition to extract fetal ECG is the commonly method.In the multi-layer wavelet,we extract the mother's peak,and use wavelet to reconstruct the mother's ECG.In the end,the fetal ECG signal is obtained by subtracting the mother's ECG from the original ECG.However,this method can't effectively remove the interference noise,and easily lead to missing fetal QRS.Using the method of least squares support vector machine,we firstly need to know the time-frequency characteristics of fetal ECG and its noise characteristics.Based on the good characteristics of EMD decomposition processing non-stationary signals,the original ECG is decomposed.On the other hand,we need to know the basics theory of least squares support vector machine.We need to use the EMD method to decompose mother's chest ECG data and send it to support vector machine for training and learning.The mother's abdominal heart signal is predicted from the sample data and then fetal ECG is obtained by subtracting the mother's heart signal from the original ECG.By preprocessing,decomposing,predicting ECG signals,We can write algorithm code to extract fetal ECG in MATLAB simulation platform from Analog ECG,international ECG database,clinical ECGExperiments show that the method based on EMD can deal with non-stationary signals better,and the method of least squares support vector machine can predict mother's abdominal electrocardiogram faster and better.We analyze the experimental results about the ECG signals from the two aspects of visualization effect and signal-to-noise ratio for various sources.Finally,it shows that the method can accurately extract the fetal ECG signals and can provide greater prenatal diagnosis.The value of medical information can be used as a clinical study of fetal heart rate.