Variable step-size LMS algorithm for estimation of time-varying and frequency-selective channels

A new strategy is proposed for deriving a variable step-size least mean squares (VSS-LMS) algorithm suited for the estimation of time-varying and frequency-selective communications channels modelled as wide sense stationary channels-uncorrelated scattering (WSS-US). The variable step size is obtained by minimising the analytical expression of the mean square weight estimation error as a function of time. Computer simulation results are presented which show that the new VSS-LMS algorithm so obtained produces performance gains in relation to the conventional LMS algorithm and for some channel models may also outperform the recursive least squares algorithm.