Influence of low frequency noise in adaptive estimation using the LMS algorithm

Adaptive systems are employed in the cancelation of noises and estimation of periodic and quasiperiodic signals. Amongst these signals are the electrocardiogram (ECG), impedance cardiography (ZCG), brain evoked potentials and modulated signals in telecommunication applications. In this paper we study the behavior of the weights of the LMS algorithm when used to estimate the coefficients of the discrete Fourier transform (DFT) of a signal under influence of low frequencies. We show theoretically that low frequency noise affects the estimation of the weights at higher frequencies. The simulation results obtained are in agreement with theoretical results. Moreover, we exemplify the problem with impedance cardiography (ZCG) signals.