A new non-linear recursive algorithm for obtaining the fetal electrocardiogram

The addition of a non-linear block at the output of an FIR adaptive filter improves its performance. A new recursive algorithm is proposed to increase the velocity of convergence of the learning schemes based on gradient descent. The behavior of the entire system is analyzed in a non-invasive extraction of fetal ECG. Experiments with real and simulated registers show that the new algorithm is better than the classic LMS for recovering fetal ECG.