Independent Component Analysis Based Research On Extracting FECG

Independent Component Analysis(ICA) is a statistical and calculational technology which elicits the hidden component of random variables and signals. Compared with other methods, it represents the characteristic to find components which satisfy independent non-Guass propertites. The assumptive observation data model of ICA is the linearity or non-linearity mixture of some unknow internal variables. Not only the internal variables appears unknown, but also the mixed system is unkown. The thesis supposes that the internal variables are independent non-Guass and distills the internal independent component of them(original signals).This thesis faciliates the ICA into Foetus'Cardiogram(FECG) processing. FECG contains lots of important information about pregnant women, such as foetus mature degree, foetus position, multi-foetus, etc. It is also an effective detective method of foetus arrhythmia and acidosis inside uterus. These information is of great importance to matrixes and foetus before childbearing.. It is useful to provide credible data for later clinic diagnoses.The main emphases of this thesis could be concluded into following steps:Based on the definition and hypothesis of ICA, we firstly analyse its basic mathematic model, its basic principles, previous data processing, centering and whiting, independence criterion and its optimized principles. We also describe common algorithms of ICA. An d compares ICA with Primary Component Analysis(PCA).Secondly, we use Fast fixed-point ICA(FICA) to distill the foetus cardiogram. After detecting the multi-electric ECG from pregnant women's body with every conductance is mixed with MECG and FECG signals, FICA is used to dispose the omnibus signals and separates the MECG and FECG signals. Then we checkout the applied effect of FICA.At last, we process the distilled signals. Using the art of wavelet threshold, we de-noise the FECG signals and finally acquire excellent clean FECG signals.