Peak constrained least-squares QMF banks

This paper presents two peak constrained least-squares quadrature mirror filter (QMF) banks design methods. The proposed design methods minimize the least-squares error of the reconstruction response and the subband filter responses with respect to the ideal responses under the given peak constraints on the reconstruction error ripple and subband filter stopband ripple without explicit specification of the transition-band bandedges of the subband filters. The proposed iterative algorithms converge efficiently in all the simulations, and achieve QMF banks with better performances and lower computational complexities than that obtained by other algorithms in literature. Furthermore, design trade-off among the peak constrained reconstruction error ripple size, subband filter ripple size, and the subband filter transition-band bandwidth of the QMF banks can be obtained from the presented algorithm. Various design examples are presented to illustrate the versatilities of the proposed design methods.