Optimizing the performance of polynomial adaptive filters: making quadratic filters converge like linear filters

摘要

The correlation properties of the input vector determine the rate of convergence of the LMS algorithm for Volterra adaptive filters and are optimal when the nonlinear input terms are uncorrelated. This correspondence presents new results on the correlation properties for second-order Volterra filters and shows that when the input signal is whitened, the nonlinear terms automatically become uncorrelated.