An elastic acoustic scattering by underwater target could be mixed with other acoustic scattering components in both time and frequency domains, and the existing signal processing methods could not discriminate the elastic feature of target in the mixed status. For solving this problem, a signal separation method for elastic acoustic scattering is proposed. Based on the highlight model of target echo, the characters of the target acoustic scattering signal when the linear frequency modulation signal is transmitted, are analyzed, and a method for mapping the acoustic scattering signal of the target to a single frequency signal is proposed. Theoretical derivation shows that there is a simple linear relationship between the acoustic scattering structure of the target and the mapped result, then the elastic acoustic scattering signal of the target can be separated by a narrow-band filtering. Simulation results show that the correlation coefficient between the separated target acoustic scattering and the orginal simulation signals are about 0.99, indicating that the acoustic scattering components in the simulation target echo can be wholly separated. Experimental data-processing results of the target acoustic scattering measureflent in an anechoic pool show that the mid-frequency enhancement effect can be observed in the spectrum of the separated elastic acoustic scattering, and every target acoustic scattering component can be recognized on the time-frequency distribution of separately processed target echo. There is a bowl-shape interference fringe on the angle-spectrum of the separated target elastic acoustic scattering components which is consistent with the theoretical signal’s feature of the elastic acoustic scattering, and the effectiveness of the separation method proposed is proven.