| Abstract: Millisecond pulsars (MSPs) and normal pulsars are key astrophysical laboratories for studying neutron star physics, testing theories of gravity, probing the interstellar medium, and detecting low-frequency gravitational waves through pulsar timing arrays. Despite more than 4,000 known pulsars, population synthesis studies predict a substantially larger undiscovered Galactic population, particularly at low radio frequencies. The GMRT High Resolution Southern Sky (GHRSS) survey is a blind, off-Galactic-plane pulsar survey conducted with the Giant Metrewave Radio Telescope (GMRT), designed to exploit its high sensitivity and wide bandwidth (200 MHz) at metre wavelengths.
In this work, we present significant enhancements to the GHRSS pulsar search pipeline, incorporating optimised dispersion measure (DM) planning, Fast Folding Algorithm (FFA)-based searches, and machine learning-assisted candidate classification. The modified FFT pipeline extends the DM coverage up to 250 pc cm⁻³, encompassing nearly 88% of the known pulsar population, while maintaining computational efficiency. To recover long-period and narrow-duty-cycle pulsars missed by traditional FFT methods, an FFA-based search implemented using the RIPTIDE framework is integrated into the pipeline.
Furthermore, we implement a Gaussian Hellinger Very Fast Decision Tree (GH-VFDT) machine learning classifier to automatically rank pulsar candidates, substantially reducing manual inspection load while maintaining high recall. This ML-assisted reanalysis of archival GHRSS data has yielded promising pulsar-like candidates, several of which have been re-observed using GMRT incoherent and phased-array beams for confirmation.
The GHRSS survey has so far resulted in over 30 discoveries, including MSPs, mildly recycled pulsars, and rotating radio transients. Ongoing observations aim to fill remaining sky coverage gaps and extend the survey to higher frequency bands, further improving sensitivity to faint and fast-spinning pulsars. This work demonstrates the effectiveness of combining advanced signal-processing techniques and machine learning to enhance pulsar discovery in large-scale radio surveys. |
