Fixed-point error analysis of two-channel perfect reconstructionfilter banks with perfect alias cancellation

摘要

This paper studies the effects of fixed-point arithmetic inntwo-channel perfect reconstruction (PR) filter banks. Practicalnimplementations of filter banks often require scaling of coefficientsnand differing binary word sizes to maintain dynamic range. When scalingnis used with fixed precision arithmetic, the perfect alias cancellationn(PAC) and PR constraints no longer hold. The main contribution of thisnpaper is the derivation of constraints whereby PAC is maintained, evennwhen coefficients are scaled and when the analysis and synthesis filternbanks use different binary word lengths. Once PAC is established, thenfixed-point effects on PR properties can be analyzed using standardnmethods. The theory is verified by comparing predicted and actualnreconstruction signal-to-noise ratios resulting from simulating ansymmetric wavelet transform