Designing of QMF Bank

Abstract

Quadrature mirror filter (QMF) banks have been of great interest during the past decade. These filters find application where a discrete signal is to split into a number of consecutive bands in the frequency domain, so that sub-band signal can be processed in an independent manner and sufficient compression may be achieved. Typical processing includes under sampling the sub-band signals, encoding them and transmitting over a channel. Eventually, at some point in the process, the sub-band signals should be recombined so that original signal is properly reconstructed. Typical application of such signal splitting includes sub-band coders for video signals [2], digital trans-multiplexers used in FDM/TDM conversion, and frequency domain speech scramblers. In this thesis, we look at the QMF design problem purely as a signal reconstruction problem. The channel is therefore assumed to be noiseless and the exact signal characteristics are not given. We treat the QMF design problem as a multivariable optimization over the filter coefficient. 

Keywords: Quadrature mirror, transfer functions, optimization