Design of prototype filter using windowing and linear optimization technique for the non-uniform filter banks

Abstract

Non-uniform filter banks (NUFBs) are being preferred because it divides the signal into unequal bands. Among perfect reconstruction (PR) and Nearly Perfect Reconstruction (NPR) filter banks, the later is preferred since the filter banks can be realized with minimal complexity. In this paper, a prototype filter is designed using the windowing techniques and the pass-band edge frequency (ωp) is optimized using the linear optimization technique for the design of a near perfect reconstructed filter bank. Among the different windowing techniques, Kaiser and dolph-chebyshev windows have been considered. The performance of the filter banks for each of these windowing techniques was measured in terms of Peak Reconstruction Error (PRE). These techniques are being applied on an Electro Cardio Gram (ECG) signal and their respective results are also being discussed.