Being the main source of vehicles' movement state parameters, wheel speed is central for control systems and its accuracy affects their performance. In order to attenuate the wheel speed sensor errors and reduce the research and manufacture cost, an error estimation method based on resilient back propagation (BP) is presented. The improved resilient BP neural network is applied to estimate the sensor errors. Matching algorithm is illustrated to realize the corresponding errors. Theoretical analysis and simulation results show that the proposed method can make the error less than 2×10^-4 rad, so it can effectively attenuate the sensor errors and improve the accuracy of the wheel speed signal.