Open AccessApplication of kernel PCA for foetal ECG estimation
Author(s) -
Xueyun Wei,
Wei Zheng
Publication year - 2018
Publication title -
electronics letters
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.375
H-Index - 146
ISSN - 1350-911X
DOI - 10.1049/el.2018.0071
Subject(s) - principal component analysis , signal (programming language) , pattern recognition (psychology) , kernel (algebra) , computer science , artificial intelligence , kernel principal component analysis , correlation , mathematics , kernel method , support vector machine , programming language , geometry , combinatorics
A new method of estimating the antepartum foetal electrocardiogram (ECG) signal by kernel principal component analysis (PCA) is presented. For ECG signals collected from the body surface of the pregnant woman, the powerful maternal ECG is the most PC, compared with the foetal ECG and other noises. Utilising the correlation between the maternal components in different lead ECG signals, the maternal components can be removed from the abdominal signal to obtain the foetal ECG estimation. However, it shows a strong non‐linearity between the maternal components in every collected signal due to the diversity of propagation path. Kernel PCA can be seen as a non‐linear form of PCA which can extract the non‐linear PCs from multidimensional data. Thus, it can be applied to the multiple leads ECG signals to eliminate the maternal ECG components and estimate the foetal ECG signal precisely. The effectiveness of the proposed method is verified by the real data experiment and compared with the existing work.