Accurate vehicle longitudinal velocity estimation is important for wheel slip ratio control in antilock braking systems. To overcome the problem of nonlinear tyre-road friction characteristic when designing an observer for velocity estimation, this paper presents a novel approach by using the model-based fuzzy technique. The nonlinear vehicle braking system is modelled by a Takagi-Sugeno fuzzy model first. A fuzzy observer is then constructed by using the available measurements of wheel angular velocity and braking torque with the estimated premise variables. All the possible disturbances and uncertainties are considered so that the designed observer is robust under an Hoo performance index from the disturbances to the estimation error. The design of the observer is achieved by solving a set of linear matrix inequalities. Numerical simulations on a quarter-vehicle braking model are used to validate the effectiveness of the proposed approach.