Optimal Prototype Filters for Near-Perfect-Reconstruction Cosine-Modulated Nonuniform Filter Banks with Rational Sampling Factors

摘要

In this paper, we propose a simple design method of nonuniform filter banks (NUFBs) with rational sampling factors. The analysis filters are respectively generated by cosine modulating several prototype filters with certain matching condition. By following a cosine roll-off function of the transition bands of the stretched prototype filters, the overall system is approximately power complementary and therefore possesses the near-prefect-reconstruction (NPR) property. Further, due to the use of the Parks-McClellan algorithm in the filter design, no objective function in the optimization is explicitly required and optimal prototype filters for NPR cosine-modulated NUFBs can be obtained. Design examples along with comparisons show that the resulting filter banks are of high filter quality and the proposed method is of simplicity and effectiveness.