Environment estimation for enhanced NLMS adaptation

A novel scheme for managing the convergence-controlling parameter of the normalized least-mean-squares (NLMS) adaptation algorithm to provide the optimal expected squared error in the subsequent sample is introduced. This optimization requires some knowledge of the environment in which the adaptation takes place. Consequently, an extended Kalman filter (EKF) is used to estimate a carefully chosen set of three parameters called the reduced adaptation state. As demonstrated by a number of simulations, the information supplied by three parameters is sufficient to provide an effective time-variation for the NLMS convergence-controlling parameter without significant increase in computational complexity