Abstract: The micro-motion feature can be utilized to space cone-shaped target discrimination, parameter estimation and so on. However, since the target translation influences the extraction of micro-motion feature, the translation usually needs to be compensated before feature extraction. A method for translation compensation and micro-motion feature extraction is proposed based on the time-varying autoregressive(TVAR) model in this paper. We first make an analysis on the scattering properties of the cone-shaped target, and derive the instantaneous frequency(IF) variations induced by the translation and micro-motion. Then the TVAR model is applied for the IF estimation, the ambiguity resolution and re-association are utilized to separate the estimated IF components. Finally, the target translation is compensated based on the polynomial fitting of the envelopes of the IF components to extract the micro-motion feature. Experiments based on electromagnetic computation data indicate that the proposed method is valid and accurate.