Subband adaptive filters with zero alias component constraints

Critically decimated subband adaptive filters (SBAF) require filtering across bands to cancel aliasing effects. The addition of the cross-band filters can slow down the convergence rate and increase the asymptotic mean square error. In this paper, a SBAF which constrains the parameters of the cross-band filters to suppress the aliasing is proposed. The cross-band parameters depend on the in-band filters, so that only the in-band parameters need to be adapted. The procedure to assemble the constrained SBAF is discussed in detail. The resulting expressions provide a quantitative basis for deciding which filters across the bands are important to include in the subband filter. The adaptive algorithm, and its the computational cost are discussed. Experiments show that this new SBAF have improved convergence rate and asymptotic mean square error when compared with standard SBAF approaches