Abstract: Following the analysis of the interference noise, a method based on Independent Component Analysis ( ICA ) to denoise the acoustic emission signals is presented to overcome the disadvantage of AE signals being strongly affected by the background noises, which shall affect the accuracy of positioning and judgment. Firstly, the signals are pretreated by the mixing factor analysis method, then, the observed signals are decomposed of several independent components by the improved independent component analysis algorithm. By setting the noise component to be zero according to some priori knowledge in time domain or frequency domain of the signals, the purposes of noise elimination could be achieved. Experimental results show that the independent component analysis can get a higher SNR and positioning accuracy than commonly de-noising method. It thus proves that the pretreatment methods and improved independent component analysis algorithm described in the text are effective.