for the problem that pulsar signal and noise are aliasing on the intrinsic mode functions (imf) during the empirical mode decomposition (emd), a denoising method based on emd with kurtosis test window is proposed. firstly, the starting point for imf refactoring is calculated by auto-correlation and cross-correlation. on that basis, the left and the right endpoints of the pulse part of signal from the former two imfs of the starting point for refactoring are acquired by the local kurtosis test. finally, the adaptive threshold method is used to remove noise and improve signal quality. the experimental results show that, compared with the other five methods, the proposed method has higher denoising performance, which can effectively eliminate the noises and retain the details in pulsar signal. © 2017, editorial office of systems engineering and electronics. all right reserved.