Dynamic convergence behaviour in Darwinian adaptive multi-notch filtering

Non-stationary narrowband interference occurring at several spectral locations can be removed by cascaded arrangement of adaptive second-order digital notch filters. The relative responsiveness of the adaptive algorithm to dynamic changes in interference locations, of course, a crucial matter in determining the effectiveness of this filtering approach-as is the steady-state accuracy when the interference centre frequencies remain fixed. The paper considers an adaption algorithm, drawn from the class of natural algorithms, referred to as Darwinian design. The Darwinian algorithm employs a relentless `survival of the fittest´ rule which drives toward optimal filtering by guided random perturbations of filter parameters. The chief worry is elapsed time, since `random search´ is notoriously time-consuming in general