Constrained adaptive least mean squared filters

Abstract

The possibility of designing constrained adaptive finite impulse response digital filters is investigated as motivated by a study of adaptive noise cancellation. The first constraint considered consists of a fixed angle between filter zeros and is implemented in a master-slave approach in which one of the zeros is adjusted adaptively and the others follow subject to the constraint. The second constraint considered is a linear constraint on the filter weights and is implemented by augmenting the error equation with Lagrangean multipliers. Simulation indicate that the approach is feasible.