Fetal ECG Signal Processing by Different ICA-based Algorithms

This article deals with fetal electrocardiography (fECG) processing using independent component analysis (ICA). Testing is performed on 7 synthetic recordings with a different level of signal-to-noise ratio (SNR) and the evaluation is performed on calculation of improvement SNR and root mean square error (RMSE). The experiment was based on testing multiple algorithms based on the ICA method, such as the algorithm based on kurtosis value, the algorithm based on negentropy value and the algorithm called kurtosis maximization ICA. The results showed that all ICA-based algorithms a lot improve SNR and have a low value of RMSE, which indicates that signals after filtration are almost similar to the reference signals. All three ICA-based algorithms could be used for fECG extraction, but the lowest accuracy was achieved by the algorithm called kurtosis maximization ICA.