Radar micro-Doppler (mD) provides a promising approach to parameter estimation and classification of micro-dynamic targets. The micro-motion states and characteristics of instantaneous mD frequency are investigated to free symmetric rigid bodies. They are found to take the precession motion, and thus induce complicated mD features with non-sinusoidal variation. Based on theoretical analysis of the spectral structure of mD time-frequency (TF) sequence, this paper proposes a rotation parameter estimation method for complicated micro-motions. The estimation method includes TF analysis, spectrogram processing, projection mapping and spectral estimation. The spin and precession rates of micro-dynamic targets can then be extracted from their radar echoes. Finally, the effectiveness of the proposed method is verified by Monte-Carlo simulations and further discussion.