An optimal model-matching approach to multirate FIR filter bank design.

摘要

A new model-matching approach to Quadrature Mirror Filter (QMF) bank design is presented. Matching the QMF bank to be designed with an ideal system forms an error system. The optimization involves FIR synthesis filters designed to minimize the reconstruction error between the QMF bank and the ideal system. Norms are used to quantify the reconstruction error. The convex nature of this problem guarantees global optimality of the resulting minimum.;While originally intended for QMF bank design, this model-matching approach is easily extended to other multirate systems, such as A/D hybrid QMF banks and trans-multiplexers. Furthermore, constraints such as linear-phase filters, cosine-modulated filters, and alias-free systems are also accommodated by the model-matching approach. In addition to the cost introduced by these constraints, tradeoffs between reconstruction properties and filter length or system delay are examined.