An Efficient Method for Designing of Modulated Filter Banks with Causal-Stable IIR Filters

摘要

This paper proposes an efficient method for the design of a class of causal-stable infinite impulse response (IIR) cosine modulated filter banks (CMFBs) and a class of IIR modified discrete Fourier transform filter banks (MDFT FBs) with a very low system delay. Instead of designing the frequency magnitude responses of the IIR prototype filters, the power spectrums of the desired filters are designed so as to reformulate the design problems as a quasi-convex optimization problem which can be solved by CVX software. When the solutions are found, the IIR prototype filters can be obtained by a spectral factorization technique. By using a bisection search, IIR FBs obtained are nearly perfect reconstruction (NPR) and which can be used as effective initial guess to some constrained nonlinear optimizer for further optimization to design perfect reconstruction (PR) IIR FBs with a very low system delay.