Research Of Separation Technology Of Fetal ECG Signal Based On ICA Separation

Electrocardiogram(ECG)is an important basis for detecting cardiac health in clinical practice.Non-invasive fetal heart rate information obtained from fetal ECG is of great significance for monitoring fetal health during pregnancy.In recent years,with the development of portable sensor technology,It has been possible to continuously detect fetal ECG through remote monitoring equipment.For example,the use of Wireless Body Sensor Networks(WBSN)to collect abdomen ECG signals to monitor the fetus on a daily basis can timely detect fetal congenital hypoxia,heart rate Disorders and other health problems for early treatment.The remote monitoring equipment gets the wearer’s physiological signal through the sensor,and uploads it to the terminal for observation and diagnosis by medical staff.The technologies required to be used mainly include: ECG signal data compression,denoising,separation and identification of ECG signal components,eigenwave detection,etc.The existing research has made many achievements in these technologies,but further development is still needed in practical application.This paper is committed to the study of obtaining fetal electrocardiogram(FECG)from abdomen electrocardiogram(AECG)obtained by remote monitoring equipment.The main work is the separation and identification of ECG signal components and the application of data compression in ECG signals.The use of non-invasive methods to separate FECG from AECG is very important for fetal health monitoring.However,due to the low signal-to-noise ratio of FECG in AECG,and including power frequency interference,baseline drift,electromyography High-amplitude maternal electrocardiogram(MECG),so the analysis of abdomen fetal ECG is still a difficult problem.The work done in this paper is as follows:1.ECG signal data compression.The ECG monitor will generate a large amount of ECG data,which brings difficulties to the data transmission and power consumption of the wearable device.The compression of the ECG data is essential.The compressed sensing theory proposes that the sampling rate of the signal is no longer limited by the bandwidth of the signal,and the sparse representation according to the structure and characteristics of the signal can greatly reduce the amount of data transmitted and stored in the ECG signal.In this paper,compressed sensing is applied to the compression of ECG signals,and two methods of generating over-complete dictionaries are compared.The first is to find a generating function with an approximate shape according to the time-domain characteristics of the ECG signal,and then expand and contract the generating function.A fixed DCT over-complete dictionary obtained by translation and discretization; the second is an over-complete dictionary obtained by using the collected ECG signal as training data and using the dictionary training algorithm KSVD algorithm.The characteristics of the electrical signal can realize the sparse decomposition of the ECG signal,and the dictionary generated based on the KSVD algorithm training has better performance.2.Abdomen FECG extraction.First,denoise the abdomen signal to eliminate noise such as baseline drift,and then perform Independent Component Analysis(ICA)preprocessing on the signal:de-averaging,whitening and other processes to reduce the workload of the algorithm; then according to the composition of AECG The characteristics of the signal in the paper expound the feasibility of applying the ICA model to FECG extraction,introduce three common objective functions in the ICA algorithm,and describe in detail the principle of choosing non-Gaussian as the measure of signal independence in this paper; Two common problems that hinder automatic processing and solve them:use the threshold discrimination method to solve the problem of polarity inversion of the separated ECG signal in the extraction of the ICA algorithm,and combine the sample entropy to solve the problem that the sequence of the separated signals cannot be determined by the ICA algorithm.In this way,the stability of the entire ECG signal processing process is ensured,and MECG is correctly identified without manual confirmation; the mother’s ECG QRS complex is eliminated by weighted singular value decomposition(SVD),and the extraction of FECG is completed..3.QRS complex detection.The QRS complex is the waveform with the largest amplitude in the ECG cycle,and the peak of the R wave is the most prominent.After the R wave is detected,the entire QRS complex can be located and the position of other waveforms can be determined.In this paper,a simple and real-time threshold detection method is used to detect QRS complexes,and the problem that it is susceptible to noise interference is improved: for the non-stationary characteristics of the ECG signal,it is easily caused by the pregnant woman’s own reasons and the influence of equipment.In the case of unstable ECG signal amplitude,sudden drop in signal amplitude,and local amplitude lower than normal value,a morphology-based ECG signal R-wave detection algorithm is proposed,which corrects the missed detection of dynamic threshold detection algorithm.Affected low-amplitude ECG R waves can be better identified.