Fast converging adaptive filter with modified SPR condition

A two-stage adaptive stochastic algorithm is introduced which modifies the strict positive real (SPR) condition used as a sufficient condition for convergence by existing algorithms. This algorithm uses a white-noise dither signal and a debiasing parameter to guarantee the convergence when the passivity condition fails and to reduce the bias in the estimated parameters without a priori information about the unknown model. The proposed algorithm is then applied to the problems of sinusoidal detection, adaptive line enhancement, and spectral estimation. The simulation results show that the proposed algorithm compares favorably with several previously published algorithms.<>