The modified Pisarenko's algorithm for adaptive spectrum estimation

The authors present a computationally efficient gradient-type algorithm for adaptively implementing Pisarenko's method. Using the property that the Z-transform of the error filter equation has linear phase, they adaptively implement Pisarenko's method with the weight vector p(n) whose dimension is (L+1), instead of the weight vector W(n) whose dimension is (2L+1), where L is the number of sinusoids buried in noise. The required computational complexity per iteration is about half that of previous implementations of Pisarenko's method. Simulation results are presented. The authors derive the convergence properties of the modified Pisarenko's algorithm in the vicinity of a stationary point. Additionally, they derive an expression for the range of the feedback coefficient for stable convergence.<
>