In order to identify the ground battlefield armored vehicle target through passive acoustic recognition, select representative objectives include two kinds of tanks and two kinds of crawler armored vehicle as the noise acquisition, use the signal of truck and thunder as interference term, the noise signal are preweighted, framed, calculated the power spectrum and then entered the Mel filter bank, using mean value of noise signal’s MFCC as eigenvalue in order to construct feature vector, with support vector machine as classifier, establish an armored vehicle classification method which classification rate can reach more than 95%. Research indicates that this method is adaptable to the classification of tank and armored vehicle and can effectively resist the impact of noncombat target’s noise signal on battlefield, provided right information for battlefield decisions.