Aiming at the problem of radar measurement system tracking accuracy degradation affected by wild values, this paper proposes an Expectation Maximization Modified Unbiased Converted Measurement Kalman Filter (EM-MUCMKF). Firstly, the target measurement information is performed an unbiased conversion. Then a time prediction and a measurement update are performed under the Kalman filter framework. Next the updated target state is regarded as a new measurement to perform an unbiased conversion again, for the covariance revaluated at the updated target state is more accurate and much less noisy. Finally, under the framework of expectation maximization, the adaptive factor matrix of the measurement noise covariance is calculated which is used to correct the measurement noise covariance. The simulation results show that compared with traditional algorithms, the proposed algorithm can get more accurate target state estimation in the environment affected by wild values.