In view of the singular value problem of acceleration random decrement (RD) signature and the extraction problem of structural real mode shapes in the practice of modal identification, a corresponding signature truncation method and a minimum distance method are proposed to improve the modal identification effect. Firstly, the reason of the singular value of acceleration RD signature is analyzed, and the signature truncation method is proposed for the subsequent processing with the Ibrahim time domain technology. Then, by using an optimization model, the computation formula of the minimum distance method is established to facilitate the extraction of structural real mode shapes from the identified complex mode shapes. Finally, the numerical example and model experiment are applied to verify the feasibility of the proposed methods. The results show that a rational value of the truncation ratio for the acceleration RD signature would be 1/200~1/50, and the minimum distance method is applicable with high noise resistance for extracting real mode shapes under all kinds of responses.