Efficient design of a class of multiplier-less perfect reconstruction two-channel filter banks and wavelets with prescribed output accuracy

Abstract

This paper proposes a novel algorithm for the design and hardware reduction of a class of multiplier-less two-channel perfect reconstruction (PR) filter banks (FB) using sum-of-powers-of-two (SOPOT) coefficients. It minimizes a more realistic hardware cost, such as adder cells, subject to a prescribed output accuracy taking account of rounding and overflow effects, instead of using just the SOPOT terms as in conventional methods. Furthermore, by using multiplier blocks (MB) to implement the filters in the FBs, significant overall saving in hardware resources can be achieved. An effective random search algorithm is also proposed to solve the design problem, which is also applicable to PR IIR FBs with highly nonlinear objective functions. Perfect reconstruction multirate filter have important applications in signal analysis, signal coding and the design of wavelet bases.