SNR estimation for X-ray pulsar based on wavelet-aided random forest

For the X-ray pulsar navigation system, the signal-to-noise rate (SNR) of the accumulative pulsar profile is an important index. The inaccurate setting of SNR affects the performance of Kalman filter in the X-ray pulsar navigation system seriously. Thus, the accurate estimation of SNR for the X-ray pulsar profile is crucial. However, the research of the SNR estimation is still blank. To fill this blank, we propose the SNR estimation method for X-ray pulsar. This method comprehensively uses wavelet and random forest to realize the SNR estimation, and includes training and testing. The profile set consists of the X-ray pulsar profile with different SNRs. In training model, the accumulative pulsar profile is decomposed by wavelet. As the wavelet low frequency coefficients contain the main energy of the accumulative pulsar profile, it is extracted to be used as the input of random forest. We train random forest to get the model. In testing model, we decompose accumulative pulsar profile by wavelet, and use wavelet low frequency coefficients as the input of the trained random forest model to estimate the SNR of the accumulative pulsar profile. Simulation results demonstrate that the accuracy of the SNR estimation reaches 0.7dB, and fulfill the requirement for the X-ray pulsar navigation system.