A stable fast affine projection adaptation algorithm suitable for low-cost processors

Abstract

The fast affine projection (FAP) adaptation algorithm possesses a superior convergence property, but most of its implementations so far suffer from instability resulting from error propagation and accumulation in a matrix inversion process. A stable yet simple FAP algorithm is proposed, where the matrix inversion is performed based on an optimization scheme in the field of non-linear programming. Proven to be stable and immune to numerical errors, the new algorithm is suitable for a wide range of voice applications.