Abstract: The greatest advantage of ultrasound is that it causes no destruction when used to identify the metal materials.Since the single expression ability of characteristic value is poor,we have analyzed the echo signal in the time domain and frequency domain,and integrated the characteristics of the two signals as anti-fake identification information.To realize automatic identification classification of metal,three kinds of metal materials have been taken for samples in this experiment.The recognition correct rate can reach 93.3% by using binary-tree architecture classifier which is constructed with twin support vectormachines(TSVM).Results show that the method has a certain practical value,and it provides a feasible method for metal identification.