The unbiased gradient type LS algorithm for adaptive spectrum estimation

A gradient-type-least-squares (LS) algorithm for adaptive implementation of Pisarenko's method is presented. The algorithm, which uses the Lagrange multiplier technique for parameter updating, has faster convergence and improved tracking capability when compared to a least-mean-squares-type algorithm. Moreover, since the gradient-type LS algorithm does not use an approximation for deriving the updated weight vector, tracking errors in the transient region are smaller than with Reddy's LS algorithm. Convergence analysis of the gradient-type LS algorithm in the vicinity of a stationary point shows that it is unbiased.<
>