As the traditional Compressive Sensing(CS) based Micro-Doppler(M-D) feature extraction methods cannot be utilized in wideband radar, the internal property of returned signals induced by the M-D targets is focused, and a modified Multiple Measurement Vectors(MMV) model is constructed based on the Linear Frequency Modulated(LFM) signal. Combining the frequency estimation approaches and the Orthogonal Matching Pursuit(OMP) algorithm, the sparse solution of the MMV model is solved, and therefore the M-D features are obtained. Simulation results indicate that, the proposed approach outperforms traditional Single Measurement Vector(SMV) based methods in the robustness, especially under noisy conditions.