Fetal ECG extraction using adaptive functional link artificial neural network

Abstract

In this paper, a nonlinear adaptive noise canceller (ANC) based on the functional link artificial neural network (FLANN) is proposed for extracting fetal electrocardiogram (FECG). The FLANN is placed in parallel with an FIR filter. The two filters are designated to implement the linear and nonlinear mappings between the maternal ECG (MECG) and the composite abdominal ECG (AECG) acquired in the thoracic and abdominal areas, respectively. The AECG is used as the primary signal while the MECG serves as the reference signal in the ANC. The FLANN is essentially a linear combiner with nonlinear input, and thus enjoys many nice properties such as fast convergence, computational efficiency etc. The LMS algorithm is applied to the proposed ANC. Application to a real dataset reveals that the proposed system is quite effective and outperforms previous ANC with only FIR filters.