This paper presents a design methodology for the optimum linear filter and predictor applied to robot sensor signals, as well as a sensitivity analysis of the kalman algorithms for uncertainties in the estimation of signal and noise parameters. Simulation of the filtration and prediction processes was made assuming a first-order spectrum of a pure signal and white measurement noise. Calculations of the algorithm errors, dependent on the accuracy of the signal and noise parameter estimation, were done for various spectra of the signal and for various signal-to-noise ratios. Furthermore, the sensitivity curves of the Kalman filter and predictor are presented. The outlined considerations might be helpful for designers when synthesizing optimum linear digital filters applied to sensor signals. Although that particular procedure has been designed for a robot tactile sensor application, the conclusions are more general, and applications may be found elsewhere.