Abstract: X-ray pulsar navigation is a deep space exploration technology with promising development potential. Its navigation accuracy is primarily affected by the precision of time of arrival of X-ray pulsar signal. X-ray pulsar signal denoising technology is expected to provide strong signal support for X-ray pulsar navigation. A key challenge in X-ray pulsar signal denoising is to retain the detailed information of X-ray pulsar signals while effectively suppressing noise. In empirical mode decomposition (EMD) threshold denoising, the number of aliasing intrinsic mode functions (IMFs), threshold functions and thresholds are main factors affecting the denoising effect. This paper analyzes IMF components using fast fourier transform (FFT) and selects noisy high-frequency mixed IMFs based on the frequency domain sparsity of IMFs. For the problem of choosing threshold functions and thresholds, a composite evaluation index (CEI) is proposed to identify the optimal combination of threshold function and threshold estimation method. The effectiveness of the proposed method is verified by numerical simulation. Simulation and test results indicate that this method may have application prospects in pulsar navigation.