Abstract: In order to solve the impact of the impulse noise in the universe on the long time epoch folding of X-ray pulsar profile, an innovative denoising method using the wavelet packet transformation is proposed. First, the system model containing noisy signal for X-ray pulsar is established. The time series of X-ray pulsar are sampled as certain time interval and folded as its rotational period. Secondly, the signal transformed by wavelet packet is computed according to the given entropy, aiming to obtain the optimum basis. Finally, the impulse noise is filtered with the proper threshold. The experimental results show that the impulse noise can be filtered by the proposed method, and the algorithm improves the signal-to-noise ratio (SNR) by about 40% compared with other methods under low SNR conditions.