In order to obtain the longitudinal speed of the in-wheel motor driven vehicle, a new estimation algorithm for the extended Kalman filter was designed based on signals of wheel speed and vehicle body acceleration. First, the discrete state equation and measurement equation of the research object were established. Then, two extended Kalman filters (EKFs), including a noise filter and an estimation filer, were designed to deal with measuring signals and estimate the vehicle's longitudinal speed, respectively. Finally, the parameters obtained by the estimation filer were adjusted through the fuzzy controller to ensure the adaptivity of the algorithm. The simulation results show that the error between the estimated speed and the actual speed was less than 2% when the road adhesion coefficient was 1.00, and the error was less than 10% when the road adhesion coefficient was 0.25.