Compressed Sensing Based RFI Mitigation and Restoration for Pulsar Signals

In pulsar signal processing, two primary difficulties are (1) radio-frequency interference (RFI) mitigation and (2) information loss due to preprocessing and mitigation itself. Linear mitigation methods have a difficulty in RFI modeling, and accommodate a limited range of RFI morphologies. Thresholding methods suffer from manual factors and adaptability. There is also a distinct lack of methods dedicated to information loss. In this paper, a novel method “CS-Pulsar” is proposed. It carries out compressed sensing (CS) on time-frequency signals to accomplish RFI mitigation and signal restoration simultaneously. Curvelets allow an optimal sparse representation for multichannel pulsar signals containing the time-of-arrival dispersion relationship. CS-Pulsar mitigation is implemented using a regularized least-squares framework that does not require the statistics of RFI to be known beforehand. CS-Pulsar implements channel restoration, and useful signal contents are retrieved from the measurement error by a morphological component analysis aided by the root-mean-square envelope. These two steps allow CS-Pulsar to provide key signal details for special astrophysical purposes. Experiments of signal restoration for pulsar data from the Nanshan 26 m radio telescope reveal the advantage of CS-Pulsar. The method successfully removes false peaks due to on-pulse RFI in multipeaked pulsar profiles. CS-Pulsar also increases the timing accuracy and signal-to-noise ratio proving its feasibilities and prospects in astrophysical measurements.