Abstract: This paper describes a method for extracting the aerodynamic derivatives of a helicopter from flight test data by means of frequency technique, Kalman filtering and Bayesian maximum likelihood technique. The method can be sum-maried as follow:1.The high frequency effects from the rotor are eliminated using the frequency technique.2.By using the Kalman filter, the random noise can be reduced considerably.3.By using maximum likelihood technique, the random noise is minimized, and the final derivatives are extracted.By using this method, the bias error can be minimized. The results show that the multiple correlation coefficient would approach 1. 0, and latent roots would approach true values. The results also show that the present method is more accurate than any other techniques, for example, the least square technique, frequency technique, Kalman filtering technique, improved Kalman filtering technique or the maximum likelihood technique.