A novel spectral factorization method and its application in the design of filter banks and wavelets

Abstract

One of the difficulties encountered in spectral factorization is achieving computational accuracy when computing a spectral factor whose zeros are on the unit circle and/or the order is high. An iterative approach is presented for computing the spectral factor accurately. It is shown that the new method performs well and can be used in the design of filter banks satisfying perfect reconstruction and orthonormal compactly supported wavelets. Design examples are included.<>