A transform domain LMS algorithm with an adaptive step size equation

A transform domain least mean-square (TDLMS) adaptive algorithm employing a new adaptive step size control technique is proposed. The variable step size approximates a theoretical optimal one based on a proposed cost function. As a result, considerable improvement in convergence speed is attained in early stages of convergence with small misadjustment near the optimum. Simulation experiments clearly demonstrate the enhanced convergence characteristics of the proposed algorithm.