Micro-Doppler effect is of great potential for radar target recognition since the micro-dynamics of structures on the target can be detected with it. The recognition process includes time-frequency analysis of the returned radar signal, feature extraction from time-frequency distribution images, and classification according to the feature set. In this paper, the time-frequency distribution images of four different types of micro-dynamics, which are vibration, rotation, coning and tumbling, are given and analyzed first. Then a method is proposed for feature extraction from time-frequency distribution images of micro-Doppler dynamics. Simulated experimental results have shown that high classification performances for different classifiers have been achieved