Performance analysis of the variable step-size Griffiths' LMS algorithm for fetal electrocardiography

This paper evaluates the performance of the variable step-size Griffiths” LMS (VGLMS) algorithm for the separation of the fetal electrocardiogram (FECG) from the maternal electrocardiogram (MECG). The VGLMS algorithm is known for its observation noise free step size and gradient and hence its efficiency is explored for the purpose of fetal electrocardiography. Its performance is analyzed and compared with that of the normalized least mean squares (NLMS) algorithm. The comparison of these adaptive algorithms is carried out by choosing optimal filter lengths and step sizes which are chosen in order to maximize the signal to noise ratios (SNR) of the filtered signals. The SNR achieved by the algorithm, the power spectral density (PSD) and the weight vector convergence are chosen as the parameters of comparison. The superiority of the VGLMS algorithm over the NLMS algorithm is witnessed. The average SNRs achieved were 4.129dB and 0.922dB by the VGLMS and NLMS algorithms respectively.