Design of subband filter banks with implementation constraints

Abstract

An important consideration during filter bank design is the implementational issue. Since actual implementations use finite precision, some filter bank properties such as perfect reconstruction no longer remains valid at finite precision. A property is structural if it is valid even with such approximation of values during implementation. In this work we develop novel lifting based algorithms for design of 2 channel FIR filter banks that are structurally perfect reconstruction and structurally linear phase. Further, other implementation constraints addressed in this project are structurally imposed vanishing moment (regularity) of the filter bank, more efficient implementation of a filter bank (by reduction of number of coefficients and their dynamic range), and factorization based recursive design as well as realization of filter bank.