Abstract: The enormous noise existing in the pipeline leakage acoustic emission detection would greatly influence the precision of localization of leakage sources. Aiming at this problem, this paper firstly used EMD to decompose the pipeline leakage acoustic emission signal. Then, the high frequency was de-noised by wavelet analysis in order to retain the useful signal. Finally, the useful signal was selected to reconstruct the acoustic emission signal and the leakage sources location was determined by correlation analysis. The experimental results showed that, compared with the single processing method, the location accuracy of the leakage detection was increased by using EMD and wavelet analysis.