Abstract
A block affine projection algorithm that is less complex but mathematically equivalent to a recently proposed Gauss-Seidel pseudo affine projection (GS-PAP) algorithm is proposed. This reduced complexity is achieved by applying a partitioning method to the original sample-by-sample algorithm. It is shown that the derived algorithm has better convergence and tracking abilities than those of the NLMS algorithm. Its application in an acoustic echo cancellation is investigated.