OPTIMAL t UADRATURE MIRROR FILTER BANK DESIGN

Abstract

Quadrature mirror filters have been used extensively for splitting speech/image signals into sub-bands which result in no aliasing problem when the sub-band signals are decimated and then interpolated and recombined. QMF bank is basically a two-channel filter bank consists of analysis filters, down samplers at transmission end, and the receiving end consists of up-samplers, & synthesis filters. The objective is to design analysis and synthesis filters to achieve perfect reconstruction. QMF banks suffer from different types of distortions like Amplitude distortion, phase distortion and aliasing distortion. In this dissertation, alias-free QMF bank has been assumed and phase distortion is eliminated by considering linear phase FIR filters. So, the design of two-channel linear-phase quadrature mirror filter (QMF) banks constructed by real finite impulse response (FIR) digital filters has been considered. In this report, various design methods have been reviewed, and a method has been modified to improve the performance. The design examples and results of these methods are discussed in chapter 6. It has been observed that the proposed modified method is superior to that of others. Some of these methods use reconstruction error as objective and other performance metrics as constraints. Other methods involve minimization of stop-band & maximization of the pass-band energies of the filters as objective and reconstruction error as constraint. In the proposed modified method, the perfect reconstruction condition is formed in the time-domain and instead of minimizing the objective function directly, an iterative procedure is used. At the end, an application of sub-band coding for image compression is presented using the designed QMF bank. The designs and simulations are carried out using MATLAB.