Modified kernel regression method for the denoising of X-ray pulsar profiles

In this paper, a modified kernel regression algorithm is proposed to reduce the noise of pulsar profiles autonomously. Taking advantage of the classical autonomous kernel regression the presented algorithm based on the second-order derivative compensation is developed to improve the performance of the Nadaraya-Watson kernel estimator. The periodic extension technique is introduced to eliminate the boundary issue inherent in kernel regression means. Four indexes are utilized to explore the performance of the proposed method via both emulated and real data. Additionally, other widely accepted denoising methods based on wavelet transformation and empirical model decomposition are simulated to make a comparison. The experimental results have shown that the proposed algorithm achieves a higher quality profile than the compared methods, which will help to discover the emission mechanism of pulsars. According to our experiments, we would like to point out that a smooth profile cannot guarantee an accurate measurement (time of arrival, TOA), which indicates that it is not necessary to denoise the epoch folding profile for estimating the TOA information in X-ray pulsar-based navigation systems.