A robust extended Kalman filter (REKF) algorithm for the fault tolerant estimation of parameters of electro-mechanical actuator (EMA) in the presence of measurement faults is proposed. The proposed REKF is based on the evaluation of the posterior probability of the normal operation of the system, given for the current measurement. This probability is proposed to calculate via the posterior probability density of the normalised innovation sequence at the current estimation step. As a result, faults in the estimation system are corrected by the system, without affecting the good estimation behaviour. The developed REKF algorithm is applied for the parameter identification process of an EMA. The performance of the proposed filter is tested for the different types of measurement faults. The single actuator sensor faults are considered. The proposed REKF algorithm is tested for different measurement malfunction scenarios; continuous bias at measurements, measurement noise increment, instantaneous abnormal measurements and fault of zero output.