Combination of a FIR filter with a Genetic algorithm for the extraction of a fetal ECG
Malika Kedir Talha, Mohamed Saïd Guettouche, A. Bousbia-Salah · 2010
The detection and analysis of a fetal cardiac signal are the primary goals of the electronics in fetal monitoring. Adaptive Wiener filtering and quadratic error minimization such as RLMS (Recursive Least Mean Square) and NLMS (Normalized Least Mean) methods can give a satisfactory but non-optimal solution. In this paper, we propose to apply an adaptive filtering by combining a time-varying finite impulse filter (FIR) with a genetic algorithm (GA). With this solution, we can obtain the filter's coefficients which minimize the quadratic error and guarantee convergence towards the optimal filter. In order to show the impact of GA compared to the filters of Wiener, RLMS and NLMS, we realized on the same real signal recorded on mother (MECG), the extraction of the cardiac signal of fetus (FECG). A GA of eight bits and ten iterations only seems to be a filter of quality compared to the other filters.