Probabilistic Design of Long Error Free Fixed-point Polynomial Predictors and Differentiators

V.S. Dimitrov (Canada) and J.M.A. Tanskanen (Finland)


Under ordinary coefficient quantization, prediction and/orImage Processing and Applications,i


In this paper, a method for designing long fixed-point polynomial FIR predictors (FPFP) and polynomial-pre dictive FIR differentiators (FPPFD) is proposed. Our method yields filters that perform exact prediction and/or differentiation even with short coefficient word lengths. differentiation capabilities degrade, or may be totally lost. Here the filters are designed so that the prediction and/or differentiation properties are exactly preserved in fixed point implementations. Earlier it has been possible to de sign short round-off error free FPFPs and FPPFDs. In this paper, the problem is shown solvable also for long FPFPs and FPPFDs by a probabilistic algorithm.

Important Links:

Go Back