Smart sensors often require that embedded estimators are robust and blind for given averaging horizons. This brief proposes a new receding horizon (RH) finite impulse response (FIR) velocity estimator that fits these needs by utilizing data from ${N}$ recent discrete position measurements with fading weights. The conventional Kalman estimator typically exhibits poor performance and may even diverge under imprecisely defined noise statistics and/or numerical errors. In contrast, the proposed weighted RH FIR estimator does not require any information about noise, which makes it more robust and blind for a given ${N}$ . The weighted RH FIR estimator minimizes the effects of uncertainties caused by imprecisely defined noise statistics and/or numerical errors and demonstrates better robustness than the existing FIR estimators. We also discuss how to choose the optimal horizon size for the weighted RH FIR estimator. The better performance of the proposed weighted RH FIR estimator against the Kalman and FIR estimators is shown through simulations under diverse operation conditions.