Study On FECG Extraction Based On Independent Component Analysis Method And Mathematical Morphology Filtering

Pregnancy of fetal monitoring or monitoring is an important method to evaluateintrauterine safety and developmental status, the fetal ECG signal is the source of fetalheart activity signal, the signal is compared with the cardiac signals and heart soundsignals, morereflect the whole picture of the fetal heart activity, and when the fetalabnormal fetal ECG morphological changes occur earlier than the heart, and heartsounds, and more sensitive. The common method is to extract the fetal ECG signalmeasured from the maternal abdominal surface ECG, the major difficulties of thismethod is that the fetal ECG signal itself is a very weak signal, and the magnitude of thematernal ECG much larger than its amplitude, the conventional signal extractionalgorithm to extract the fetal ECG is neither clear it is difficult to identify. At home andabroad extensive algorithm wavelet transform and independent component analysis,adaptive filtering, noise cancellation technology, algorithm.In order to meet the requirements of real-time, first select the minimum meansquare error gradient algorithm for adaptive filtering algorithm, the reasons for usingthis algorithm is the algorithm to calculate the amount of small, fast convergence. Inorder to verify the separation efficiency of the algorithm, we first proposed an analogsynthesis of Maternal and Child aliasing signal through the use of the actual adult ECGsignal as the blueprint to produce simulated the FECG the model. Fetal ECG signal andadult ECG signal by comparing the similarities and differences between an adult ECGsignal analog synthesizer FECG signal. The purpose of the FECG the simulation as thesource signal, to evaluate the extraction efficiency of different signal processingtechniques, provide an important reference for the follow-up to the actual extraction ofthe FECG signal.Matlab simulation experiments show that the adaptive noise cancellationtechnology separation of mother and child signal, less computation, and the algorithmconverges faster, but the actual separation is less effective, the maternal ECG signalinterference still exists, compared with the source signal, signal reconstruction of theroot mean square error, the more obvious differences with the source signal, when theanalog EMG interference signal stack by adding noise, the signal extracted completelysubmerged in the simulated EMG noise, followed by the actual the fetal ECG signalextraction, was found, the adaptive noise cancellation algorithm failed extracted FECG signal; independent component separation technology convergence slower, but theseparation of fetal ECG is more pure, and the difference of the source signal is verysmall, root mean square error is smaller, the extraction of simulated fetal ECG signal atthe same time effective to go in addition to the interference of maternal ECG. The mostcommonly used three independent component separation algorithm model separation offetal ECG and found a common problem that can not effectively remove the baselinedrift in the independent component signal. Analysis of the reasons may be due tobaseline drift signal and ECG are homologous signal, whose independence is poor, andtherefore does not meet the basic conditions of the independent component.To effectively remove the interference of the fetal ECG baseline drift, the morepure fetal ECG, opening and closing cascade morphological filtering algorithm,construct different structural elements of the simulated fetal heart electrical signal forprocessing, to extract the appropriate baseline, and direct subtraction, the baseline tosubtract from the signal source aliasing, when elements of the structure closest to theshape of the fetal baseline, filtered baseline effect the most excellent, the root meansquare error is smaller, lower dispersion of the signal. But the use of fetal ECG signalafter the algorithm to filter out the baseline drift in the base line is almost a flat line, thelow frequency components present in the normal fetal ECG also be filtered out, thereason may be in the baseline used in the removal process is a direct subtractiontechnique, which led to the generation of this case.In order to filter out baseline drift at the same time preserving usefullow-frequency signal, this paper uses a smooth filtering, adaptive noise cancellationalgorithm. Baseline and source-aliasing signal extracted at the same time as the adaptivenoise canceller two inputs, select the appropriate iteration parameters and the step size,the separation of the analog signal. Matlab simulation experiments, confirmed that thesmaller root mean square error of the algorithm simulation isolated fetal heart signal andsource signal, and can significantly improve the separation of signal to noise ratio. Truefetal ECG signal processing show that the algorithm can effectively extract the purefetal ECG signal.