A blind extraction method of fetal electrocardiogram signal based on MNCMD-NLBCA

Fetal electrocardiogram signal extraction has very important clinical significance. Currently, most existing fetal electrocardiogram signal extraction algorithms are designed for ideal positive definite linear instantaneous mixed models, but the actual collected observation signals often do not meet the ideal standards. Therefore, the extraction results often have many problems such as inaccurate cycles, low correct recognition rates, resource waste, and time consumption. This article proposes an underdetermined nonlinear bounded component analysis method. Firstly, the observed signal is processed using multivariate nonlinear chirp mode decomposition, and then the signal is reconstructed to convert the original underdetermined situation into positive definite; Then, Gaussian technique is used for nonlinear compensation to convert the nonlinear model into a linear model; Finally, the signal is extracted using the bounded component analysis method, which has more relaxed constraint assumptions and does not require signal sources to be independent of each other. This article investigates the role of underdetermined nonlinear bounded component analysis methods in extracting fetal electrocardiogram signals. The experimental dataset mainly uses MIT-BIH, ADFECGD, and PCCDB. The experimental results show that the average similarity coefficient, MSE and SIR values of the algorithm are 0.969, 0.023 and 12.85, respectively, which proves that it can quickly and effectively extract clearer fetal electrocardiogram signals.