Study On Extraction Method Of The Fetal Electrocardiogram

The extraction of fetal electrocardiogram(FECG) is an important monitoring items inperinatal period. The purpose of the extraction is finding out health status of the fetal duringpregnancy(growth degree of the fetal, position of fetus etc). Generally speaking, placeingsome electrodes on abdominal of maternal's right position is to measure the fetalelectrocardiogram. However, signal we measured always superposes some interference andnoise, the most important of which is maternal ECG(MECG). In addition, the noise of powerline interference, respiration, electromyography is common.Many other have been reported, such as, coherence average, autocorrelation andcross-correlation, singularity value decomposition method, and principal component analysis,etc. However, these ways of extraction of FECG is subjected to much limit, such as , complexcalculation or much manual intervention.Independent component analysis(ICA) is a developing signal processing technique withthe development of the blind signal separation in recent 10 years. It has been a hot spot insignal processing field because of algorithm simple, low computational complexity and wideapplication range. In the problems of extraction of FECG, the noises of measured values aremixed, and mixing coefficient is determined by physical geometryfactor and conductivity ofsoft tissue, so they are basically constant. And these sources are independent, which accordwith conditions of ICA. So it is feasible to extract of FECG using this technology of ICA.In this paper, the common signal processing technology has been first introduced,including their characteristics, the scope of application, and contrast to their advantages anddisadvantages,then we make simulation tests. The simulation results is non-ideal. Then thetheoretical foundation of ICA that are necessary to understand and master ICA technique hasbeen introduced, including the statistics theory and information theory. Finally, the basicprinciple of ICA has been introduced. Two famous ICA algorithms, extended Infomax algorithm and fast fixed-point algorithm,have been introduced. Both algorithms have been applied to the problem of FECG extraction.The results are ideal. From these results,we can found that not only ICA algorithms canextract much clearer FECG , but also they are non-sensitive to the electrodes position. So theyare more practical.Another the innovative point of this paper is that the improved algorithm has beenproposed, it is an algorithm specially for FECG extraction, which is based on jointmaximization of kurtosis and autocorrelation.It utilizes the FECG's independence on othersources and FECG's special temporal property fully, so it can extract clearer FECG comparedto existing algorithms, the more important is it can greatly reduces noises interference andneed less human interaction.